Upload 34 files

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Fabio Sarracino - enemyx.net
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

__init__.py

@@ -0,0 +1,116 @@
+# Created by Fabio Sarracino
+__version__ = "1.3.0"
+__author__ = "Fabio Sarracino"
+__title__ = "VibeVoice ComfyUI"
+
+import logging
+import os
+import sys
+import subprocess
+
+# Setup logging
+logger = logging.getLogger("VibeVoice")
+logger.propagate = False
+
+if not logger.handlers:
+    handler = logging.StreamHandler()
+    formatter = logging.Formatter('[VibeVoice] %(message)s')
+    handler.setFormatter(formatter)
+    logger.addHandler(handler)
+    logger.setLevel(logging.INFO)
+
+def apply_timm_compatibility_patches():
+    """Apply compatibility patches for timm package conflicts"""
+    try:
+        import timm.data
+        
+        # Patch missing functions that cause import errors
+        patches = {
+            'ImageNetInfo': lambda: type('ImageNetInfo', (), {'__init__': lambda self: None})(),
+            'infer_imagenet_subset': lambda class_to_idx: 'imagenet',
+            'get_imagenet_subset_labels': lambda *args, **kwargs: [],
+            'get_imagenet_subset_info': lambda *args, **kwargs: {},
+            'resolve_data_config': lambda *args, **kwargs: {}
+        }
+        
+        for attr_name, patch_func in patches.items():
+            if not hasattr(timm.data, attr_name):
+                if attr_name == 'ImageNetInfo':
+                    setattr(timm.data, attr_name, type('ImageNetInfo', (), {'__init__': lambda self: None}))
+                else:
+                    setattr(timm.data, attr_name, patch_func)
+        
+        return True
+    except Exception as e:
+        return False
+
+def check_embedded_vibevoice():
+    """Check if embedded VibeVoice is available"""
+    vvembed_path = os.path.join(os.path.dirname(__file__), 'vvembed')
+    if not os.path.exists(vvembed_path):
+        logger.error(f"Embedded VibeVoice not found at {vvembed_path}")
+        return False
+    
+    # Add vvembed to path if not already there
+    if vvembed_path not in sys.path:
+        sys.path.insert(0, vvembed_path)
+    
+    logger.info("Using embedded VibeVoice (MIT licensed)")
+    return True
+
+def ensure_dependencies():
+    """Ensure required dependencies are installed"""
+    try:
+        import transformers
+        from packaging import version
+        if version.parse(transformers.__version__) < version.parse("4.44.0"):
+            logger.warning("Transformers version < 4.44.0, some features may not work correctly")
+    except ImportError:
+        logger.warning("Transformers not installed. Please install: pip install transformers>=4.44.0")
+        return False
+    
+    # Apply timm patches if needed
+    apply_timm_compatibility_patches()
+    
+    return True
+
+# Initialize node mappings
+NODE_CLASS_MAPPINGS = {}
+NODE_DISPLAY_NAME_MAPPINGS = {}
+
+# Register text loading node (always available)
+try:
+    from .nodes.load_text_node import LoadTextFromFileNode
+    NODE_CLASS_MAPPINGS["LoadTextFromFileNode"] = LoadTextFromFileNode
+    NODE_DISPLAY_NAME_MAPPINGS["LoadTextFromFileNode"] = "VibeVoice Load Text From File"
+except Exception as e:
+    logger.error(f"Failed to register LoadTextFromFile node: {e}")
+
+# Register VibeVoice nodes (using embedded VibeVoice)
+if check_embedded_vibevoice() and ensure_dependencies():
+    try:
+        from .nodes.single_speaker_node import VibeVoiceSingleSpeakerNode
+        from .nodes.multi_speaker_node import VibeVoiceMultipleSpeakersNode
+        from .nodes.free_memory_node import VibeVoiceFreeMemoryNode
+        
+        # Single speaker node
+        NODE_CLASS_MAPPINGS["VibeVoiceSingleSpeakerNode"] = VibeVoiceSingleSpeakerNode
+        NODE_DISPLAY_NAME_MAPPINGS["VibeVoiceSingleSpeakerNode"] = "VibeVoice Single Speaker"
+        
+        # Multi speaker node
+        NODE_CLASS_MAPPINGS["VibeVoiceMultipleSpeakersNode"] = VibeVoiceMultipleSpeakersNode
+        NODE_DISPLAY_NAME_MAPPINGS["VibeVoiceMultipleSpeakersNode"] = "VibeVoice Multiple Speakers"
+        
+        # Free memory node
+        NODE_CLASS_MAPPINGS["VibeVoiceFreeMemoryNode"] = VibeVoiceFreeMemoryNode
+        NODE_DISPLAY_NAME_MAPPINGS["VibeVoiceFreeMemoryNode"] = "VibeVoice Free Memory"
+        
+        logger.info("VibeVoice nodes registered successfully")
+        
+    except Exception as e:
+        logger.error(f"Failed to register VibeVoice nodes: {e}")
+        logger.info("Please ensure transformers>=4.44.0 is installed")
+else:
+    logger.warning("VibeVoice nodes unavailable - check embedded module and dependencies")
+
+__all__ = ['NODE_CLASS_MAPPINGS', 'NODE_DISPLAY_NAME_MAPPINGS', '__version__']

examples/Multiple-Speaker.json

@@ -0,0 +1 @@
+{"id":"e5ca15c5-18b5-4d37-8852-795692a14b29","revision":0,"last_node_id":28,"last_link_id":41,"nodes":[{"id":20,"type":"Note","pos":[-76.50189208984375,752.0235595703125],"size":[415,88],"flags":{},"order":0,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["Use Load Text From File if you want to use a .txt file instead of text-area. You can load .txt files from ComfyUI/input, ComfyUI/output or ComfyUI/temp directories."],"color":"#432","bgcolor":"#653"},{"id":16,"type":"PreviewAudio","pos":[914.1022338867188,189.5508270263672],"size":[270,88],"flags":{},"order":6,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","link":41},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null}],"outputs":[],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"PreviewAudio"},"widgets_values":[],"color":"#323","bgcolor":"#535"},{"id":15,"type":"LoadAudio","pos":[-12.263749122619629,190.64144897460938],"size":[270,136],"flags":{},"order":1,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"COMBO","widget":{"name":"audio"},"link":null},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null},{"localized_name":"upload","name":"upload","type":"AUDIOUPLOAD","widget":{"name":"upload"},"link":null}],"outputs":[{"localized_name":"AUDIO","name":"AUDIO","type":"AUDIO","links":[39]}],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"LoadAudio"},"widgets_values":["Voice1.mp3",null,null],"color":"#2a363b","bgcolor":"#3f5159"},{"id":17,"type":"LoadAudio","pos":[-11.774602890014648,403.2247009277344],"size":[270,136],"flags":{},"order":2,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"COMBO","widget":{"name":"audio"},"link":null},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null},{"localized_name":"upload","name":"upload","type":"AUDIOUPLOAD","widget":{"name":"upload"},"link":null}],"outputs":[{"localized_name":"AUDIO","name":"AUDIO","type":"AUDIO","links":[40]}],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"LoadAudio"},"widgets_values":["Voice2.mp3",null,null],"color":"#2a363b","bgcolor":"#3f5159"},{"id":19,"type":"LoadTextFromFileNode","pos":[-8.260560989379883,651.4061279296875],"size":[270,58],"flags":{},"order":3,"mode":4,"inputs":[{"localized_name":"file","name":"file","type":"COMBO","widget":{"name":"file"},"link":null}],"outputs":[{"localized_name":"text","name":"text","type":"STRING","links":null}],"properties":{"Node name for S&R":"LoadTextFromFileNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["No text files found in any directory"],"color":"#323","bgcolor":"#535"},{"id":28,"type":"VibeVoiceMultipleSpeakersNode","pos":[392.798095703125,188.42286682128906],"size":[400,388],"flags":{},"order":5,"mode":0,"inputs":[{"localized_name":"speaker1_voice","name":"speaker1_voice","shape":7,"type":"AUDIO","link":39},{"localized_name":"speaker2_voice","name":"speaker2_voice","shape":7,"type":"AUDIO","link":40},{"localized_name":"speaker3_voice","name":"speaker3_voice","shape":7,"type":"AUDIO","link":null},{"localized_name":"speaker4_voice","name":"speaker4_voice","shape":7,"type":"AUDIO","link":null},{"localized_name":"text","name":"text","type":"STRING","widget":{"name":"text"},"link":null},{"localized_name":"model","name":"model","type":"COMBO","widget":{"name":"model"},"link":null},{"localized_name":"attention_type","name":"attention_type","type":"COMBO","widget":{"name":"attention_type"},"link":null},{"localized_name":"free_memory_after_generate","name":"free_memory_after_generate","type":"BOOLEAN","widget":{"name":"free_memory_after_generate"},"link":null},{"localized_name":"diffusion_steps","name":"diffusion_steps","type":"INT","widget":{"name":"diffusion_steps"},"link":null},{"localized_name":"seed","name":"seed","type":"INT","widget":{"name":"seed"},"link":null},{"localized_name":"cfg_scale","name":"cfg_scale","type":"FLOAT","widget":{"name":"cfg_scale"},"link":null},{"localized_name":"use_sampling","name":"use_sampling","type":"BOOLEAN","widget":{"name":"use_sampling"},"link":null},{"localized_name":"temperature","name":"temperature","shape":7,"type":"FLOAT","widget":{"name":"temperature"},"link":null},{"localized_name":"top_p","name":"top_p","shape":7,"type":"FLOAT","widget":{"name":"top_p"},"link":null}],"outputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","links":[41]}],"properties":{"Node name for S&R":"VibeVoiceMultipleSpeakersNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["[1]: Hello, this is the first speaker.\n[2]: Hi there, I'm the second speaker.\n[1]: Nice to meet you!\n[2]: Nice to meet you too!","VibeVoice-Large","auto",true,20,42,"fixed",1.3,false,0.95,0.95],"color":"#223","bgcolor":"#335"},{"id":21,"type":"Note","pos":[387.95281982421875,624.7701416015625],"size":[415,88],"flags":{},"order":4,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["The first time you use a model, it will be downloaded to ComfyUI/models/vibevoice/. This can take several minutes!\nThe VibeVoice-1.5B model (about 5GB)\nThe VibeVoice-Large model (about 17GB)\nThe VibeVoice-Large-Quant-4Bit model (about 7GB)"],"color":"#432","bgcolor":"#653"}],"links":[[39,15,0,28,0,"AUDIO"],[40,17,0,28,1,"AUDIO"],[41,28,0,16,0,"AUDIO"]],"groups":[],"config":{},"extra":{"ds":{"scale":1.1,"offset":[200.3891510142744,-69.52646130728805]}},"version":0.4}

examples/Pause-Tag.json

@@ -0,0 +1 @@
+{"id":"b70cf6f7-8531-4faa-9843-9c963a4ba577","revision":0,"last_node_id":38,"last_link_id":49,"nodes":[{"id":21,"type":"Note","pos":[-128.1415557861328,534.7645263671875],"size":[415,88],"flags":{},"order":0,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["Use Load Text From File if you want to use a .txt file instead of text-area. You can load .txt files from ComfyUI/input, ComfyUI/output or ComfyUI/temp directories."],"color":"#432","bgcolor":"#653"},{"id":28,"type":"LoadTextFromFileNode","pos":[-65.6552963256836,428.2049865722656],"size":[289.5152282714844,58],"flags":{},"order":1,"mode":4,"inputs":[{"localized_name":"file","name":"file","type":"COMBO","widget":{"name":"file"},"link":null}],"outputs":[{"localized_name":"text","name":"text","type":"STRING","links":null}],"properties":{"Node name for S&R":"LoadTextFromFileNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["No text files found in any directory"],"color":"#323","bgcolor":"#535"},{"id":15,"type":"LoadAudio","pos":[15.256911277770996,126.44892883300781],"size":[270,136],"flags":{},"order":2,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"COMBO","widget":{"name":"audio"},"link":null},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null},{"localized_name":"upload","name":"upload","type":"AUDIOUPLOAD","widget":{"name":"upload"},"link":null}],"outputs":[{"localized_name":"AUDIO","name":"AUDIO","type":"AUDIO","links":[48]}],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"LoadAudio"},"widgets_values":["Voice.mp3",null,null],"color":"#2a363b","bgcolor":"#3f5159"},{"id":16,"type":"PreviewAudio","pos":[892.3655395507812,127.41075897216797],"size":[270,88],"flags":{},"order":6,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","link":49},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null}],"outputs":[],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"PreviewAudio"},"widgets_values":[],"color":"#323","bgcolor":"#535"},{"id":22,"type":"Note","pos":[365.11663818359375,535.5830078125],"size":[415,88],"flags":{},"order":3,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["The first time you use a model, it will be downloaded to ComfyUI/models/vibevoice/. This can take several minutes!\nThe VibeVoice-1.5B model (about 5GB)\nThe VibeVoice-Large model (about 17GB)\nThe VibeVoice-Large-Quant-4Bit model (about 7GB)"],"color":"#432","bgcolor":"#653"},{"id":38,"type":"Note","pos":[818.8140869140625,270.3061218261719],"size":[415,88],"flags":{},"order":4,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["[pause]: add 1 second of silence.\n[pause:{number}] add {number}ms of pause\nWARNING: the pause tag forces the text to be split into chunks. This may worsen the model’s ability to understand the context. The model’s context is represented ONLY by its own chunk."],"color":"#432","bgcolor":"#653"},{"id":37,"type":"VibeVoiceSingleSpeakerNode","pos":[376.42669677734375,126.94989013671875],"size":[400,352],"flags":{},"order":5,"mode":0,"inputs":[{"localized_name":"voice_to_clone","name":"voice_to_clone","shape":7,"type":"AUDIO","link":48},{"localized_name":"text","name":"text","type":"STRING","widget":{"name":"text"},"link":null},{"localized_name":"model","name":"model","type":"COMBO","widget":{"name":"model"},"link":null},{"localized_name":"attention_type","name":"attention_type","type":"COMBO","widget":{"name":"attention_type"},"link":null},{"localized_name":"free_memory_after_generate","name":"free_memory_after_generate","type":"BOOLEAN","widget":{"name":"free_memory_after_generate"},"link":null},{"localized_name":"diffusion_steps","name":"diffusion_steps","type":"INT","widget":{"name":"diffusion_steps"},"link":null},{"localized_name":"seed","name":"seed","type":"INT","widget":{"name":"seed"},"link":null},{"localized_name":"cfg_scale","name":"cfg_scale","type":"FLOAT","widget":{"name":"cfg_scale"},"link":null},{"localized_name":"use_sampling","name":"use_sampling","type":"BOOLEAN","widget":{"name":"use_sampling"},"link":null},{"localized_name":"temperature","name":"temperature","shape":7,"type":"FLOAT","widget":{"name":"temperature"},"link":null},{"localized_name":"top_p","name":"top_p","shape":7,"type":"FLOAT","widget":{"name":"top_p"},"link":null},{"localized_name":"max_words_per_chunk","name":"max_words_per_chunk","shape":7,"type":"INT","widget":{"name":"max_words_per_chunk"},"link":null}],"outputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","links":[49]}],"properties":{"Node name for S&R":"VibeVoiceSingleSpeakerNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["Hello, this is a test of the VibeVoice text-to-speech system. [pause] Do you like my voice? [pause:500] What's your name?","VibeVoice-1.5B","auto",true,20,42,"fixed",1.3,false,0.95,0.95,250],"color":"#223","bgcolor":"#335"}],"links":[[48,15,0,37,0,"AUDIO"],[49,37,0,16,0,"AUDIO"]],"groups":[],"config":{},"extra":{"ds":{"scale":1.1000000000000005,"offset":[127.54923408733805,16.966619865757746]}},"version":0.4}

examples/Single-Speaker.json

@@ -0,0 +1 @@
+{"id":"c6ef8963-032c-45f6-954f-b5f6b354343b","revision":0,"last_node_id":37,"last_link_id":49,"nodes":[{"id":21,"type":"Note","pos":[-128.1415557861328,534.7645263671875],"size":[415,88],"flags":{},"order":0,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["Use Load Text From File if you want to use a .txt file instead of text-area. You can load .txt files from ComfyUI/input, ComfyUI/output or ComfyUI/temp directories."],"color":"#432","bgcolor":"#653"},{"id":28,"type":"LoadTextFromFileNode","pos":[-65.6552963256836,428.2049865722656],"size":[289.5152282714844,58],"flags":{},"order":1,"mode":4,"inputs":[{"localized_name":"file","name":"file","type":"COMBO","widget":{"name":"file"},"link":null}],"outputs":[{"localized_name":"text","name":"text","type":"STRING","links":null}],"properties":{"Node name for S&R":"LoadTextFromFileNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["No text files found in any directory"],"color":"#323","bgcolor":"#535"},{"id":37,"type":"VibeVoiceSingleSpeakerNode","pos":[376.42669677734375,126.94989013671875],"size":[400,352],"flags":{},"order":4,"mode":0,"inputs":[{"localized_name":"voice_to_clone","name":"voice_to_clone","shape":7,"type":"AUDIO","link":48},{"localized_name":"text","name":"text","type":"STRING","widget":{"name":"text"},"link":null},{"localized_name":"model","name":"model","type":"COMBO","widget":{"name":"model"},"link":null},{"localized_name":"attention_type","name":"attention_type","type":"COMBO","widget":{"name":"attention_type"},"link":null},{"localized_name":"free_memory_after_generate","name":"free_memory_after_generate","type":"BOOLEAN","widget":{"name":"free_memory_after_generate"},"link":null},{"localized_name":"diffusion_steps","name":"diffusion_steps","type":"INT","widget":{"name":"diffusion_steps"},"link":null},{"localized_name":"seed","name":"seed","type":"INT","widget":{"name":"seed"},"link":null},{"localized_name":"cfg_scale","name":"cfg_scale","type":"FLOAT","widget":{"name":"cfg_scale"},"link":null},{"localized_name":"use_sampling","name":"use_sampling","type":"BOOLEAN","widget":{"name":"use_sampling"},"link":null},{"localized_name":"temperature","name":"temperature","shape":7,"type":"FLOAT","widget":{"name":"temperature"},"link":null},{"localized_name":"top_p","name":"top_p","shape":7,"type":"FLOAT","widget":{"name":"top_p"},"link":null},{"localized_name":"max_words_per_chunk","name":"max_words_per_chunk","shape":7,"type":"INT","widget":{"name":"max_words_per_chunk"},"link":null}],"outputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","links":[49]}],"properties":{"Node name for S&R":"VibeVoiceSingleSpeakerNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["Hello, this is a test of the VibeVoice text-to-speech system.","VibeVoice-1.5B","auto",true,20,42,"fixed",1.3,false,0.95,0.95,250],"color":"#223","bgcolor":"#335"},{"id":16,"type":"PreviewAudio","pos":[892.3655395507812,127.41075897216797],"size":[270,88],"flags":{},"order":5,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","link":49},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null}],"outputs":[],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"PreviewAudio"},"widgets_values":[],"color":"#323","bgcolor":"#535"},{"id":15,"type":"LoadAudio","pos":[15.256911277770996,126.44892883300781],"size":[270,136],"flags":{},"order":3,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"COMBO","widget":{"name":"audio"},"link":null},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null},{"localized_name":"upload","name":"upload","type":"AUDIOUPLOAD","widget":{"name":"upload"},"link":null}],"outputs":[{"localized_name":"AUDIO","name":"AUDIO","type":"AUDIO","links":[48]}],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"LoadAudio"},"widgets_values":["Voice.mp3",null,null],"color":"#2a363b","bgcolor":"#3f5159"},{"id":22,"type":"Note","pos":[365.11663818359375,535.5830078125],"size":[415,88],"flags":{},"order":2,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["The first time you use a model, it will be downloaded to ComfyUI/models/vibevoice/. This can take several minutes!\nThe VibeVoice-1.5B model (about 5GB)\nThe VibeVoice-Large model (about 17GB)\nThe VibeVoice-Large-Quant-4Bit model (about 7GB)"],"color":"#432","bgcolor":"#653"}],"links":[[48,15,0,37,0,"AUDIO"],[49,37,0,16,0,"AUDIO"]],"groups":[],"config":{},"extra":{"ds":{"scale":1.1000000000000005,"offset":[249.36741590551986,40.60298350212133]}},"version":0.4}

examples/VibeVoice-Unload-Memory.json

@@ -0,0 +1 @@
+{"id":"fc471b7e-ccef-427f-be3f-29dec93a90ea","revision":0,"last_node_id":37,"last_link_id":47,"nodes":[{"id":34,"type":"VibeVoiceFreeMemoryNode","pos":[913.2552490234375,126.25599670410156],"size":[189.03964233398438,26],"flags":{},"order":6,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","link":47}],"outputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","links":[42]}],"properties":{"Node name for S&R":"VibeVoiceFreeMemoryNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":[],"color":"#322","bgcolor":"#533"},{"id":35,"type":"Note","pos":[809.6192016601562,208.98324584960938],"size":[415,88],"flags":{},"order":0,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["The VibeVoice Free Memory node releases memory as soon as it receives the audio input (acting as a passthrough for the audio itself). In this specific use case, however, it’s redundant, since it would be enough to enable the “free_memory_after_generate” parameter of the previous node. The ideal use case is, for example, when you have a loop generating multiple audio clips, and only after the final generation you pass the last audio and free the memory."],"color":"#432","bgcolor":"#653"},{"id":21,"type":"Note","pos":[-128.1415557861328,534.7645263671875],"size":[415,88],"flags":{},"order":1,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["Use Load Text From File if you want to use a .txt file instead of text-area. You can load .txt files from ComfyUI/input, ComfyUI/output or ComfyUI/temp directories."],"color":"#432","bgcolor":"#653"},{"id":16,"type":"PreviewAudio","pos":[1271.0958251953125,126.20075988769531],"size":[270,88],"flags":{},"order":7,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","link":42},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null}],"outputs":[],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"PreviewAudio"},"widgets_values":[],"color":"#323","bgcolor":"#535"},{"id":15,"type":"LoadAudio","pos":[15.256911277770996,126.44892883300781],"size":[270,136],"flags":{},"order":2,"mode":0,"inputs":[{"localized_name":"audio","name":"audio","type":"COMBO","widget":{"name":"audio"},"link":null},{"localized_name":"audioUI","name":"audioUI","type":"AUDIO_UI","widget":{"name":"audioUI"},"link":null},{"localized_name":"upload","name":"upload","type":"AUDIOUPLOAD","widget":{"name":"upload"},"link":null}],"outputs":[{"localized_name":"AUDIO","name":"AUDIO","type":"AUDIO","links":[46]}],"properties":{"cnr_id":"comfy-core","ver":"0.3.49","Node name for S&R":"LoadAudio"},"widgets_values":["Voice.mp3",null,null],"color":"#2a363b","bgcolor":"#3f5159"},{"id":37,"type":"VibeVoiceSingleSpeakerNode","pos":[353.6889343261719,125.9052505493164],"size":[400,328],"flags":{},"order":5,"mode":0,"inputs":[{"localized_name":"voice_to_clone","name":"voice_to_clone","shape":7,"type":"AUDIO","link":46},{"localized_name":"text","name":"text","type":"STRING","widget":{"name":"text"},"link":null},{"localized_name":"model","name":"model","type":"COMBO","widget":{"name":"model"},"link":null},{"localized_name":"attention_type","name":"attention_type","type":"COMBO","widget":{"name":"attention_type"},"link":null},{"localized_name":"free_memory_after_generate","name":"free_memory_after_generate","type":"BOOLEAN","widget":{"name":"free_memory_after_generate"},"link":null},{"localized_name":"diffusion_steps","name":"diffusion_steps","type":"INT","widget":{"name":"diffusion_steps"},"link":null},{"localized_name":"seed","name":"seed","type":"INT","widget":{"name":"seed"},"link":null},{"localized_name":"cfg_scale","name":"cfg_scale","type":"FLOAT","widget":{"name":"cfg_scale"},"link":null},{"localized_name":"use_sampling","name":"use_sampling","type":"BOOLEAN","widget":{"name":"use_sampling"},"link":null},{"localized_name":"temperature","name":"temperature","shape":7,"type":"FLOAT","widget":{"name":"temperature"},"link":null},{"localized_name":"top_p","name":"top_p","shape":7,"type":"FLOAT","widget":{"name":"top_p"},"link":null}],"outputs":[{"localized_name":"audio","name":"audio","type":"AUDIO","links":[47]}],"properties":{"Node name for S&R":"VibeVoiceSingleSpeakerNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["Hello, this is a test of the VibeVoice text-to-speech system.","VibeVoice-1.5B","auto",false,20,42,"fixed",1.3,false,0.95,0.95],"color":"#223","bgcolor":"#335"},{"id":28,"type":"LoadTextFromFileNode","pos":[-65.6552963256836,428.2049865722656],"size":[289.5152282714844,58],"flags":{},"order":3,"mode":4,"inputs":[{"localized_name":"file","name":"file","type":"COMBO","widget":{"name":"file"},"link":null}],"outputs":[{"localized_name":"text","name":"text","type":"STRING","links":null}],"properties":{"Node name for S&R":"LoadTextFromFileNode","cnr_id":"VibeVoice-ComfyUI","ver":"5a24489a7b0bf0c406d291dd51e82a085d338d44"},"widgets_values":["No text files found in any directory"],"color":"#323","bgcolor":"#535"},{"id":22,"type":"Note","pos":[349.11663818359375,500.03680419921875],"size":[415,88],"flags":{},"order":4,"mode":0,"inputs":[],"outputs":[],"properties":{},"widgets_values":["The first time you use a model, it will be downloaded to ComfyUI/models/vibevoice/. This can take several minutes!\nThe VibeVoice-1.5B model (about 5GB)\nThe VibeVoice-Large model (about 17GB)\nThe VibeVoice-Large-Quant-4Bit model (about 7GB)"],"color":"#432","bgcolor":"#653"}],"links":[[42,34,0,16,0,"AUDIO"],[46,15,0,37,0,"AUDIO"],[47,37,0,34,0,"AUDIO"]],"groups":[],"config":{},"extra":{"ds":{"scale":1,"offset":[186.3110536351836,148.09475114094386]}},"version":0.4}

node_list.json

@@ -0,0 +1,6 @@
+{
+  "VibeVoice Load Text From File": "Load .txt from ComfyUI input/output/temp",
+  "VibeVoice Single Speaker": "Single-speaker TTS with optional voice cloning",
+  "VibeVoice Multiple Speakers": "Multi-speaker TTS ([1]..[4]) with optional clones",
+  "VibeVoice Free Memory": "Frees loaded VibeVoice models; passthrough audio"
+}

nodes/__init__.py

@@ -0,0 +1,17 @@
+# Created by Fabio Sarracino
+# Nodes module for VibeVoiceWrapper
+"""
+This module contains all the ComfyUI nodes for VibeVoice integration.
+"""
+
+from .load_text_node import LoadTextFromFileNode
+from .single_speaker_node import VibeVoiceSingleSpeakerNode
+from .multi_speaker_node import VibeVoiceMultipleSpeakersNode
+from .free_memory_node import VibeVoiceFreeMemoryNode
+
+__all__ = [
+    'LoadTextFromFileNode', 
+    'VibeVoiceSingleSpeakerNode', 
+    'VibeVoiceMultipleSpeakersNode',
+    'VibeVoiceFreeMemoryNode'
+]

nodes/base_vibevoice.py

@@ -0,0 +1,872 @@
+# Created by Fabio Sarracino
+# Base class for VibeVoice nodes with common functionality
+
+import logging
+import os
+import tempfile
+import torch
+import numpy as np
+import re
+import gc
+from typing import List, Optional, Tuple, Any
+
+# Setup logging
+logger = logging.getLogger("VibeVoice")
+
+# Import for interruption support
+try:
+    import execution
+    INTERRUPTION_SUPPORT = True
+except ImportError:
+    INTERRUPTION_SUPPORT = False
+    logger.warning("Interruption support not available")
+
+# Check for SageAttention availability
+try:
+    from sageattention import sageattn
+    SAGE_AVAILABLE = True
+    logger.info("SageAttention available for acceleration")
+except ImportError:
+    SAGE_AVAILABLE = False
+    logger.debug("SageAttention not available - install with: pip install sageattention")
+
+def get_optimal_device():
+    """Get the best available device (cuda, mps, or cpu)"""
+    if torch.cuda.is_available():
+        return "cuda"
+    elif hasattr(torch.backends, 'mps') and torch.backends.mps.is_available():
+        return "mps"
+    else:
+        return "cpu"
+
+def get_device_map():
+    """Get device map for model loading"""
+    device = get_optimal_device()
+    # Note: device_map "auto" might work better for MPS in some cases
+    return device if device != "mps" else "mps"
+
+class BaseVibeVoiceNode:
+    """Base class for VibeVoice nodes containing common functionality"""
+    
+    def __init__(self):
+        self.model = None
+        self.processor = None
+        self.current_model_path = None
+        self.current_attention_type = None
+    
+    def free_memory(self):
+        """Free model and processor from memory"""
+        try:
+            if self.model is not None:
+                del self.model
+                self.model = None
+            
+            if self.processor is not None:
+                del self.processor
+                self.processor = None
+            
+            self.current_model_path = None
+            
+            # Force garbage collection and clear CUDA cache if available
+            import gc
+            gc.collect()
+            
+            if torch.cuda.is_available():
+                torch.cuda.empty_cache()
+                torch.cuda.synchronize()
+            
+            logger.info("Model and processor memory freed successfully")
+            
+        except Exception as e:
+            logger.error(f"Error freeing memory: {e}")
+    
+    def _check_dependencies(self):
+        """Check if VibeVoice is available and import it with fallback installation"""
+        try:
+            import sys
+            import os
+            
+            # Add vvembed to path
+            current_dir = os.path.dirname(os.path.abspath(__file__))
+            parent_dir = os.path.dirname(current_dir)
+            vvembed_path = os.path.join(parent_dir, 'vvembed')
+            
+            if vvembed_path not in sys.path:
+                sys.path.insert(0, vvembed_path)
+            
+            # Import from embedded version
+            from modular.modeling_vibevoice_inference import VibeVoiceForConditionalGenerationInference
+            
+            logger.info(f"Using embedded VibeVoice from {vvembed_path}")
+            return None, VibeVoiceForConditionalGenerationInference
+            
+        except ImportError as e:
+            logger.error(f"Embedded VibeVoice import failed: {e}")
+            
+            # Try fallback to installed version if available
+            try:
+                import vibevoice
+                from vibevoice.modular.modeling_vibevoice_inference import VibeVoiceForConditionalGenerationInference
+                logger.warning("Falling back to system-installed VibeVoice")
+                return vibevoice, VibeVoiceForConditionalGenerationInference
+            except ImportError:
+                pass
+            
+            raise Exception(
+                "VibeVoice embedded module import failed. Please ensure the vvembed folder exists "
+                "and transformers>=4.51.3 is installed."
+            )
+    
+    def _apply_sage_attention(self):
+        """Apply SageAttention to the loaded model by monkey-patching attention layers"""
+        try:
+            from sageattention import sageattn
+            import torch.nn.functional as F
+            
+            # Counter for patched layers
+            patched_count = 0
+            
+            def patch_attention_forward(module):
+                """Recursively patch attention layers to use SageAttention"""
+                nonlocal patched_count
+                
+                # Check if this module has scaled_dot_product_attention
+                if hasattr(module, 'forward'):
+                    original_forward = module.forward
+                    
+                    # Create wrapper that replaces F.scaled_dot_product_attention with sageattn
+                    def sage_forward(*args, **kwargs):
+                        # Temporarily replace F.scaled_dot_product_attention
+                        original_sdpa = F.scaled_dot_product_attention
+                        
+                        def sage_sdpa(query, key, value, attn_mask=None, dropout_p=0.0, is_causal=False, scale=None, **kwargs):
+                            """Wrapper that converts sdpa calls to sageattn"""
+                            # Log any unexpected parameters for debugging
+                            if kwargs:
+                                unexpected_params = list(kwargs.keys())
+                                logger.debug(f"SageAttention: Ignoring unsupported parameters: {unexpected_params}")
+                            
+                            try:
+                                # SageAttention expects tensors in specific format
+                                # Transformers typically use (batch, heads, seq_len, head_dim)
+                                
+                                # Check tensor dimensions to determine layout
+                                if query.dim() == 4:
+                                    # 4D tensor: (batch, heads, seq, dim)
+                                    batch_size = query.shape[0]
+                                    num_heads = query.shape[1]
+                                    seq_len_q = query.shape[2]
+                                    seq_len_k = key.shape[2]
+                                    head_dim = query.shape[3]
+                                    
+                                    # Reshape to (batch*heads, seq, dim) for HND layout
+                                    query_reshaped = query.reshape(batch_size * num_heads, seq_len_q, head_dim)
+                                    key_reshaped = key.reshape(batch_size * num_heads, seq_len_k, head_dim)
+                                    value_reshaped = value.reshape(batch_size * num_heads, seq_len_k, head_dim)
+                                    
+                                    # Call sageattn with HND layout
+                                    output = sageattn(
+                                        query_reshaped, key_reshaped, value_reshaped,
+                                        is_causal=is_causal,
+                                        tensor_layout="HND"  # Heads*batch, seqN, Dim
+                                    )
+                                    
+                                    # Output should be (batch*heads, seq_len_q, head_dim)
+                                    # Reshape back to (batch, heads, seq, dim)
+                                    if output.dim() == 3:
+                                        output = output.reshape(batch_size, num_heads, seq_len_q, head_dim)
+                                    
+                                    return output
+                                else:
+                                    # For 3D tensors, assume they're already in HND format
+                                    output = sageattn(
+                                        query, key, value,
+                                        is_causal=is_causal,
+                                        tensor_layout="HND"
+                                    )
+                                    return output
+                                    
+                            except Exception as e:
+                                # If SageAttention fails, fall back to original implementation
+                                logger.debug(f"SageAttention failed, using original: {e}")
+                                # Call with proper arguments - scale is a keyword argument in PyTorch 2.0+
+                                # Pass through any additional kwargs that the original sdpa might support
+                                if scale is not None:
+                                    return original_sdpa(query, key, value, attn_mask=attn_mask, 
+                                                       dropout_p=dropout_p, is_causal=is_causal, scale=scale, **kwargs)
+                                else:
+                                    return original_sdpa(query, key, value, attn_mask=attn_mask, 
+                                                       dropout_p=dropout_p, is_causal=is_causal, **kwargs)
+                        
+                        # Replace the function
+                        F.scaled_dot_product_attention = sage_sdpa
+                        
+                        try:
+                            # Call original forward with patched attention
+                            result = original_forward(*args, **kwargs)
+                        finally:
+                            # Restore original function
+                            F.scaled_dot_product_attention = original_sdpa
+                        
+                        return result
+                    
+                    # Check if this module likely uses attention
+                    # Look for common attention module names
+                    module_name = module.__class__.__name__.lower()
+                    if any(name in module_name for name in ['attention', 'attn', 'multihead']):
+                        module.forward = sage_forward
+                        patched_count += 1
+                
+                # Recursively patch child modules
+                for child in module.children():
+                    patch_attention_forward(child)
+            
+            # Apply patching to the entire model
+            patch_attention_forward(self.model)
+            
+            logger.info(f"Patched {patched_count} attention layers with SageAttention")
+            
+            if patched_count == 0:
+                logger.warning("No attention layers found to patch - SageAttention may not be applied")
+                
+        except Exception as e:
+            logger.error(f"Failed to apply SageAttention: {e}")
+            logger.warning("Continuing with standard attention implementation")
+    
+    def load_model(self, model_name: str, model_path: str, attention_type: str = "auto"):
+        """Load VibeVoice model with specified attention implementation
+        
+        Args:
+            model_name: The display name of the model (e.g., "VibeVoice-Large-Quant-4Bit")
+            model_path: The HuggingFace model path
+            attention_type: The attention implementation to use
+        """
+        # Check if we need to reload model due to attention type change
+        current_attention = getattr(self, 'current_attention_type', None)
+        if (self.model is None or 
+            getattr(self, 'current_model_path', None) != model_path or
+            current_attention != attention_type):
+            
+            # Free existing model before loading new one (important for attention type changes)
+            if self.model is not None and (current_attention != attention_type or getattr(self, 'current_model_path', None) != model_path):
+                logger.info(f"Freeing existing model before loading with new settings (attention: {current_attention} -> {attention_type})")
+                self.free_memory()
+            
+            try:
+                vibevoice, VibeVoiceInferenceModel = self._check_dependencies()
+                
+                # Set ComfyUI models directory
+                import folder_paths
+                models_dir = folder_paths.get_folder_paths("checkpoints")[0]
+                comfyui_models_dir = os.path.join(os.path.dirname(models_dir), "vibevoice")
+                os.makedirs(comfyui_models_dir, exist_ok=True)
+                
+                # Force HuggingFace to use ComfyUI directory
+                original_hf_home = os.environ.get('HF_HOME')
+                original_hf_cache = os.environ.get('HUGGINGFACE_HUB_CACHE')
+                
+                os.environ['HF_HOME'] = comfyui_models_dir
+                os.environ['HUGGINGFACE_HUB_CACHE'] = comfyui_models_dir
+                
+                # Import time for timing
+                import time
+                start_time = time.time()
+                
+                # Suppress verbose logs
+                import transformers
+                import warnings
+                transformers.logging.set_verbosity_error()
+                warnings.filterwarnings("ignore", category=UserWarning)
+                
+                # Check if model exists locally
+                model_dir = os.path.join(comfyui_models_dir, f"models--{model_path.replace('/', '--')}")
+                model_exists_in_comfyui = os.path.exists(model_dir)
+                
+                # Check if this is a quantized model based on the model name
+                is_quantized_4bit = "Quant-4Bit" in model_name
+                is_quantized_8bit = "Quant-8Bit" in model_name  # Future support
+                
+                # Prepare attention implementation kwargs
+                model_kwargs = {
+                    "cache_dir": comfyui_models_dir,
+                    "trust_remote_code": True,
+                    "torch_dtype": torch.bfloat16,
+                    "device_map": get_device_map(),
+                }
+                
+                # Handle 4-bit quantized model loading
+                if is_quantized_4bit:
+                    # Check if CUDA is available (required for 4-bit quantization)
+                    if not torch.cuda.is_available():
+                        raise Exception("4-bit quantized models require a CUDA GPU. Please use standard models on CPU/MPS.")
+                    
+                    # Try to import bitsandbytes
+                    try:
+                        from transformers import BitsAndBytesConfig
+                        logger.info("Loading 4-bit quantized model with bitsandbytes...")
+                        
+                        # Configure 4-bit quantization
+                        bnb_config = BitsAndBytesConfig(
+                            load_in_4bit=True,
+                            bnb_4bit_compute_dtype=torch.bfloat16,
+                            bnb_4bit_use_double_quant=True,
+                            bnb_4bit_quant_type='nf4'
+                        )
+                        model_kwargs["quantization_config"] = bnb_config
+                        model_kwargs["device_map"] = "cuda"  # Force CUDA for 4-bit
+                        model_kwargs["subfolder"] = "4bit"  # Point to 4bit subfolder
+                        
+                    except ImportError:
+                        raise Exception(
+                            "4-bit quantized models require 'bitsandbytes' library.\n"
+                            "Please install it with: pip install bitsandbytes\n"
+                            "Or use the standard VibeVoice models instead."
+                        )
+                
+                # Set attention implementation based on user selection
+                use_sage_attention = False
+                if attention_type == "sage":
+                    # SageAttention requires special handling - can't be set via attn_implementation
+                    if not SAGE_AVAILABLE:
+                        logger.warning("SageAttention not installed, falling back to sdpa")
+                        logger.warning("Install with: pip install sageattention")
+                        model_kwargs["attn_implementation"] = "sdpa"
+                    elif not torch.cuda.is_available():
+                        logger.warning("SageAttention requires CUDA GPU, falling back to sdpa")
+                        model_kwargs["attn_implementation"] = "sdpa"
+                    else:
+                        # Don't set attn_implementation for sage, will apply after loading
+                        use_sage_attention = True
+                        logger.info("Will apply SageAttention after model loading")
+                elif attention_type != "auto":
+                    model_kwargs["attn_implementation"] = attention_type
+                    logger.info(f"Using {attention_type} attention implementation")
+                else:
+                    # Auto mode - let transformers decide the best implementation
+                    logger.info("Using auto attention implementation selection")
+                
+                # Try to load locally first
+                try:
+                    if model_exists_in_comfyui:
+                        model_kwargs["local_files_only"] = True
+                        logger.info(f"Loading model from local cache: {model_path}")
+                        if is_quantized_4bit:
+                            logger.info(f"Using 4-bit quantization with subfolder: {model_kwargs.get('subfolder', 'None')}")
+                        self.model = VibeVoiceInferenceModel.from_pretrained(
+                            model_path,
+                            **model_kwargs
+                        )
+                    else:
+                        raise FileNotFoundError("Model not found locally")
+                except (FileNotFoundError, OSError) as e:
+                    logger.info(f"Downloading {model_path}...")
+                    if is_quantized_4bit:
+                        logger.info(f"Downloading 4-bit quantized model with subfolder: {model_kwargs.get('subfolder', 'None')}")
+                    
+                    model_kwargs["local_files_only"] = False
+                    self.model = VibeVoiceInferenceModel.from_pretrained(
+                        model_path,
+                        **model_kwargs
+                    )
+                    elapsed = time.time() - start_time
+                else:
+                    elapsed = time.time() - start_time
+                
+                # Verify model was loaded
+                if self.model is None:
+                    raise Exception("Model failed to load - model is None after loading")
+                
+                # Load processor with proper error handling
+                from processor.vibevoice_processor import VibeVoiceProcessor
+                
+                # Prepare processor kwargs
+                processor_kwargs = {
+                    "trust_remote_code": True,
+                    "cache_dir": comfyui_models_dir
+                }
+                
+                # Add subfolder for quantized models
+                if is_quantized_4bit:
+                    processor_kwargs["subfolder"] = "4bit"
+                
+                try:
+                    # First try with local files if model was loaded locally
+                    if model_exists_in_comfyui:
+                        processor_kwargs["local_files_only"] = True
+                        self.processor = VibeVoiceProcessor.from_pretrained(
+                            model_path, 
+                            **processor_kwargs
+                        )
+                    else:
+                        # Download from HuggingFace
+                        self.processor = VibeVoiceProcessor.from_pretrained(
+                            model_path,
+                            **processor_kwargs
+                        )
+                except Exception as proc_error:
+                    logger.warning(f"Failed to load processor from {model_path}: {proc_error}")
+                    
+                    # Check if error is about missing Qwen tokenizer
+                    if "Qwen" in str(proc_error) and "tokenizer" in str(proc_error).lower():
+                        logger.info("Downloading required Qwen tokenizer files...")
+                        # The processor needs the Qwen tokenizer, ensure it's available
+                        try:
+                            from transformers import AutoTokenizer
+                            # Pre-download the Qwen tokenizer that VibeVoice depends on
+                            _ = AutoTokenizer.from_pretrained(
+                                "Qwen/Qwen2.5-1.5B",
+                                trust_remote_code=True,
+                                cache_dir=comfyui_models_dir
+                            )
+                            logger.info("Qwen tokenizer downloaded, retrying processor load...")
+                        except Exception as tokenizer_error:
+                            logger.warning(f"Failed to download Qwen tokenizer: {tokenizer_error}")
+                    
+                    logger.info("Attempting to load processor with fallback method...")
+                    
+                    # Fallback: try loading without local_files_only constraint
+                    try:
+                        self.processor = VibeVoiceProcessor.from_pretrained(
+                            model_path,
+                            local_files_only=False,
+                            trust_remote_code=True,
+                            cache_dir=comfyui_models_dir
+                        )
+                    except Exception as fallback_error:
+                        logger.error(f"Processor loading failed completely: {fallback_error}")
+                        raise Exception(
+                            f"Failed to load VibeVoice processor. Error: {fallback_error}\n"
+                            f"This might be due to missing tokenizer files. Try:\n"
+                            f"1. Ensure you have internet connection for first-time download\n"
+                            f"2. Clear the ComfyUI/models/vibevoice folder and retry\n"
+                            f"3. Install transformers: pip install transformers>=4.51.3\n"
+                            f"4. Manually download Qwen tokenizer: from transformers import AutoTokenizer; AutoTokenizer.from_pretrained('Qwen/Qwen2.5-1.5B')"
+                        )
+                
+                # Restore environment variables
+                if original_hf_home is not None:
+                    os.environ['HF_HOME'] = original_hf_home
+                elif 'HF_HOME' in os.environ:
+                    del os.environ['HF_HOME']
+                    
+                if original_hf_cache is not None:
+                    os.environ['HUGGINGFACE_HUB_CACHE'] = original_hf_cache
+                elif 'HUGGINGFACE_HUB_CACHE' in os.environ:
+                    del os.environ['HUGGINGFACE_HUB_CACHE']
+                
+                # Move to appropriate device (skip for quantized models as they use device_map)
+                if not is_quantized_4bit and not is_quantized_8bit:
+                    device = get_optimal_device()
+                    if device == "cuda":
+                        self.model = self.model.cuda()
+                    elif device == "mps":
+                        self.model = self.model.to("mps")
+                else:
+                    logger.info("Quantized model already mapped to device via device_map")
+                
+                # Apply SageAttention if requested and available
+                if use_sage_attention and SAGE_AVAILABLE:
+                    self._apply_sage_attention()
+                    logger.info("SageAttention successfully applied to model")
+                    
+                self.current_model_path = model_path
+                self.current_attention_type = attention_type
+                
+            except Exception as e:
+                logger.error(f"Failed to load VibeVoice model: {str(e)}")
+                raise Exception(f"Model loading failed: {str(e)}")
+    
+    def _create_synthetic_voice_sample(self, speaker_idx: int) -> np.ndarray:
+        """Create synthetic voice sample for a specific speaker"""
+        sample_rate = 24000
+        duration = 1.0
+        samples = int(sample_rate * duration)
+        
+        t = np.linspace(0, duration, samples, False)
+        
+        # Create realistic voice-like characteristics for each speaker
+        # Use different base frequencies for different speaker types
+        base_frequencies = [120, 180, 140, 200]  # Mix of male/female-like frequencies
+        base_freq = base_frequencies[speaker_idx % len(base_frequencies)]
+        
+        # Create vowel-like formants (like "ah" sound) - unique per speaker
+        formant1 = 800 + speaker_idx * 100  # First formant
+        formant2 = 1200 + speaker_idx * 150  # Second formant
+        
+        # Generate more voice-like waveform
+        voice_sample = (
+            # Fundamental with harmonics (voice-like)
+            0.6 * np.sin(2 * np.pi * base_freq * t) +
+            0.25 * np.sin(2 * np.pi * base_freq * 2 * t) +
+            0.15 * np.sin(2 * np.pi * base_freq * 3 * t) +
+            
+            # Formant resonances (vowel-like characteristics)
+            0.1 * np.sin(2 * np.pi * formant1 * t) * np.exp(-t * 2) +
+            0.05 * np.sin(2 * np.pi * formant2 * t) * np.exp(-t * 3) +
+            
+            # Natural breath noise (reduced)
+            0.02 * np.random.normal(0, 1, len(t))
+        )
+        
+        # Add natural envelope (like human speech pattern)
+        # Quick attack, slower decay with slight vibrato (unique per speaker)
+        vibrato_freq = 4 + speaker_idx * 0.3  # Slightly different vibrato per speaker
+        envelope = (np.exp(-t * 0.3) * (1 + 0.1 * np.sin(2 * np.pi * vibrato_freq * t)))
+        voice_sample *= envelope * 0.08  # Lower volume
+        
+        return voice_sample.astype(np.float32)
+    
+    def _prepare_audio_from_comfyui(self, voice_audio, target_sample_rate: int = 24000) -> Optional[np.ndarray]:
+        """Prepare audio from ComfyUI format to numpy array"""
+        if voice_audio is None:
+            return None
+            
+        # Extract waveform from ComfyUI audio format
+        if isinstance(voice_audio, dict) and "waveform" in voice_audio:
+            waveform = voice_audio["waveform"]
+            input_sample_rate = voice_audio.get("sample_rate", target_sample_rate)
+            
+            # Convert to numpy (handling BFloat16 tensors)
+            if isinstance(waveform, torch.Tensor):
+                # Convert to float32 first as numpy doesn't support BFloat16
+                audio_np = waveform.cpu().float().numpy()
+            else:
+                audio_np = np.array(waveform)
+            
+            # Handle different audio shapes
+            if audio_np.ndim == 3:  # (batch, channels, samples)
+                audio_np = audio_np[0, 0, :]  # Take first batch, first channel
+            elif audio_np.ndim == 2:  # (channels, samples)
+                audio_np = audio_np[0, :]  # Take first channel
+            # If 1D, leave as is
+            
+            # Resample if needed
+            if input_sample_rate != target_sample_rate:
+                target_length = int(len(audio_np) * target_sample_rate / input_sample_rate)
+                audio_np = np.interp(np.linspace(0, len(audio_np), target_length), 
+                                   np.arange(len(audio_np)), audio_np)
+            
+            # Ensure audio is in correct range [-1, 1]
+            audio_max = np.abs(audio_np).max()
+            if audio_max > 0:
+                audio_np = audio_np / max(audio_max, 1.0)  # Normalize
+            
+            return audio_np.astype(np.float32)
+        
+        return None
+    
+    def _get_model_mapping(self) -> dict:
+        """Get model name mappings"""
+        return {
+            "VibeVoice-1.5B": "microsoft/VibeVoice-1.5B",
+            "VibeVoice-Large": "aoi-ot/VibeVoice-Large",
+            "VibeVoice-Large-Quant-4Bit": "DevParker/VibeVoice7b-low-vram"
+        }
+    
+    def _split_text_into_chunks(self, text: str, max_words: int = 250) -> List[str]:
+        """Split long text into manageable chunks at sentence boundaries
+        
+        Args:
+            text: The text to split
+            max_words: Maximum words per chunk (default 250 for safety)
+        
+        Returns:
+            List of text chunks
+        """
+        import re
+        
+        # Split into sentences (handling common abbreviations)
+        # This regex tries to split on sentence endings while avoiding common abbreviations
+        sentence_pattern = r'(?<=[.!?])\s+(?=[A-Z])'
+        sentences = re.split(sentence_pattern, text)
+        
+        # If regex split didn't work well, fall back to simple split
+        if len(sentences) == 1 and len(text.split()) > max_words:
+            # Fall back to splitting on any period followed by space
+            sentences = text.replace('. ', '.|').split('|')
+            sentences = [s.strip() for s in sentences if s.strip()]
+        
+        chunks = []
+        current_chunk = []
+        current_word_count = 0
+        
+        for sentence in sentences:
+            sentence = sentence.strip()
+            if not sentence:
+                continue
+                
+            sentence_words = sentence.split()
+            sentence_word_count = len(sentence_words)
+            
+            # If single sentence is too long, split it further
+            if sentence_word_count > max_words:
+                # Split long sentence at commas or semicolons
+                sub_parts = re.split(r'[,;]', sentence)
+                for part in sub_parts:
+                    part = part.strip()
+                    if not part:
+                        continue
+                    part_words = part.split()
+                    part_word_count = len(part_words)
+                    
+                    if current_word_count + part_word_count > max_words and current_chunk:
+                        # Save current chunk
+                        chunks.append(' '.join(current_chunk))
+                        current_chunk = [part]
+                        current_word_count = part_word_count
+                    else:
+                        current_chunk.append(part)
+                        current_word_count += part_word_count
+            else:
+                # Check if adding this sentence would exceed the limit
+                if current_word_count + sentence_word_count > max_words and current_chunk:
+                    # Save current chunk and start new one
+                    chunks.append(' '.join(current_chunk))
+                    current_chunk = [sentence]
+                    current_word_count = sentence_word_count
+                else:
+                    # Add sentence to current chunk
+                    current_chunk.append(sentence)
+                    current_word_count += sentence_word_count
+        
+        # Add remaining chunk
+        if current_chunk:
+            chunks.append(' '.join(current_chunk))
+        
+        # If no chunks were created, return the original text
+        if not chunks:
+            chunks = [text]
+        
+        logger.info(f"Split text into {len(chunks)} chunks (max {max_words} words each)")
+        for i, chunk in enumerate(chunks):
+            word_count = len(chunk.split())
+            logger.debug(f"Chunk {i+1}: {word_count} words")
+        
+        return chunks
+    
+    def _parse_pause_keywords(self, text: str) -> List[Tuple[str, Any]]:
+        """Parse [pause] and [pause:ms] keywords from text
+        
+        Args:
+            text: Text potentially containing pause keywords
+            
+        Returns:
+            List of tuples: ('text', str) or ('pause', duration_ms)
+        """
+        segments = []
+        # Pattern matches [pause] or [pause:1500] where 1500 is milliseconds
+        pattern = r'\[pause(?::(\d+))?\]'
+        
+        last_end = 0
+        for match in re.finditer(pattern, text):
+            # Add text segment before pause (if any)
+            if match.start() > last_end:
+                text_segment = text[last_end:match.start()].strip()
+                if text_segment:  # Only add non-empty text segments
+                    segments.append(('text', text_segment))
+            
+            # Add pause segment with duration (default 1000ms = 1 second)
+            duration_ms = int(match.group(1)) if match.group(1) else 1000
+            segments.append(('pause', duration_ms))
+            last_end = match.end()
+        
+        # Add remaining text after last pause (if any)
+        if last_end < len(text):
+            remaining_text = text[last_end:].strip()
+            if remaining_text:
+                segments.append(('text', remaining_text))
+        
+        # If no pauses found, return original text as single segment
+        if not segments:
+            segments.append(('text', text))
+        
+        logger.debug(f"Parsed text into {len(segments)} segments (including pauses)")
+        return segments
+    
+    def _generate_silence(self, duration_ms: int, sample_rate: int = 24000) -> dict:
+        """Generate silence audio tensor for specified duration
+        
+        Args:
+            duration_ms: Duration of silence in milliseconds
+            sample_rate: Sample rate (default 24000 Hz for VibeVoice)
+            
+        Returns:
+            Audio dict with silence waveform
+        """
+        # Calculate number of samples for the duration
+        num_samples = int(sample_rate * duration_ms / 1000.0)
+        
+        # Create silence tensor with shape (1, 1, num_samples) to match audio format
+        silence_waveform = torch.zeros(1, 1, num_samples, dtype=torch.float32)
+        
+        logger.info(f"Generated {duration_ms}ms silence ({num_samples} samples)")
+        
+        return {
+            "waveform": silence_waveform,
+            "sample_rate": sample_rate
+        }
+    
+    def _format_text_for_vibevoice(self, text: str, speakers: list) -> str:
+        """Format text with speaker information for VibeVoice using correct format"""
+        # Remove any newlines from the text to prevent parsing issues
+        # The processor splits by newline and expects each line to have "Speaker N:" format
+        text = text.replace('\n', ' ').replace('\r', ' ')
+        # Clean up multiple spaces
+        text = ' '.join(text.split())
+        
+        # VibeVoice expects format: "Speaker 1: text" not "Name: text"
+        if len(speakers) == 1:
+            return f"Speaker 1: {text}"
+        else:
+            # Check if text already has proper Speaker N: format
+            if re.match(r'^\s*Speaker\s+\d+\s*:', text, re.IGNORECASE):
+                return text
+            # If text has name format, convert to Speaker N format
+            elif any(f"{speaker}:" in text for speaker in speakers):
+                formatted_text = text
+                for i, speaker in enumerate(speakers):
+                    formatted_text = formatted_text.replace(f"{speaker}:", f"Speaker {i+1}:")
+                return formatted_text
+            else:
+                # Plain text, assign to first speaker
+                return f"Speaker 1: {text}"
+    
+    def _generate_with_vibevoice(self, formatted_text: str, voice_samples: List[np.ndarray], 
+                                cfg_scale: float, seed: int, diffusion_steps: int, use_sampling: bool,
+                                temperature: float = 0.95, top_p: float = 0.95) -> dict:
+        """Generate audio using VibeVoice model"""
+        try:
+            # Ensure model and processor are loaded
+            if self.model is None or self.processor is None:
+                raise Exception("Model or processor not loaded")
+            
+            # Set seeds for reproducibility
+            torch.manual_seed(seed)
+            if torch.cuda.is_available():
+                torch.cuda.manual_seed(seed)
+                torch.cuda.manual_seed_all(seed)  # For multi-GPU
+            
+            # Also set numpy seed for any numpy operations
+            np.random.seed(seed)
+            
+            # Set diffusion steps
+            self.model.set_ddpm_inference_steps(diffusion_steps)
+            logger.info(f"Starting audio generation with {diffusion_steps} diffusion steps...")
+            
+            # Check for interruption before starting generation
+            if INTERRUPTION_SUPPORT:
+                try:
+                    import comfy.model_management as mm
+                    
+                    # Check if we're being interrupted right now
+                    # The interrupt flag is reset by ComfyUI before each node execution
+                    # So we only check model_management's throw_exception_if_processing_interrupted
+                    # which is the proper way to check for interruption
+                    mm.throw_exception_if_processing_interrupted()
+                    
+                except ImportError:
+                    # If comfy.model_management is not available, skip this check
+                    pass
+            
+            # Prepare inputs using processor
+            inputs = self.processor(
+                [formatted_text],  # Wrap text in list
+                voice_samples=[voice_samples], # Provide voice samples for reference
+                return_tensors="pt",
+                return_attention_mask=True
+            )
+            
+            # Move to device
+            device = next(self.model.parameters()).device
+            inputs = {k: v.to(device) if isinstance(v, torch.Tensor) else v for k, v in inputs.items()}
+            
+            # Estimate tokens for user information (not used as limit)
+            text_length = len(formatted_text.split())
+            estimated_tokens = int(text_length * 2.5)  # More accurate estimate for display
+            
+            # Log generation start with explanation
+            logger.info(f"Generating audio with {diffusion_steps} diffusion steps...")
+            logger.info(f"Note: Progress bar shows max possible tokens, not actual needed (~{estimated_tokens} estimated)")
+            logger.info("The generation will stop automatically when audio is complete")
+            
+            # Create stop check function for interruption support
+            stop_check_fn = None
+            if INTERRUPTION_SUPPORT:
+                def check_comfyui_interrupt():
+                    """Check if ComfyUI has requested interruption"""
+                    try:
+                        if hasattr(execution, 'PromptExecutor') and hasattr(execution.PromptExecutor, 'interrupted'):
+                            interrupted = execution.PromptExecutor.interrupted
+                            if interrupted:
+                                logger.info("Generation interrupted by user via stop_check_fn")
+                            return interrupted
+                    except:
+                        pass
+                    return False
+                
+                stop_check_fn = check_comfyui_interrupt
+            
+            # Generate with official parameters
+            with torch.no_grad():
+                if use_sampling:
+                    # Use sampling mode (less stable but more varied)
+                    output = self.model.generate(
+                        **inputs,
+                        tokenizer=self.processor.tokenizer,
+                        cfg_scale=cfg_scale,
+                        max_new_tokens=None,
+                        do_sample=True,
+                        temperature=temperature,
+                        top_p=top_p,
+                        stop_check_fn=stop_check_fn,
+                    )
+                else:
+                    # Use deterministic mode like official examples
+                    output = self.model.generate(
+                        **inputs,
+                        tokenizer=self.processor.tokenizer,
+                        cfg_scale=cfg_scale,
+                        max_new_tokens=None,
+                        do_sample=False,  # More deterministic generation
+                        stop_check_fn=stop_check_fn,
+                    )
+                
+                # Check if we got actual audio output
+                if hasattr(output, 'speech_outputs') and output.speech_outputs:
+                    speech_tensors = output.speech_outputs
+                    
+                    if isinstance(speech_tensors, list) and len(speech_tensors) > 0:
+                        audio_tensor = torch.cat(speech_tensors, dim=-1)
+                    else:
+                        audio_tensor = speech_tensors
+                    
+                    # Ensure proper format (1, 1, samples)
+                    if audio_tensor.dim() == 1:
+                        audio_tensor = audio_tensor.unsqueeze(0).unsqueeze(0)
+                    elif audio_tensor.dim() == 2:
+                        audio_tensor = audio_tensor.unsqueeze(0)
+                    
+                    # Convert to float32 for compatibility with downstream nodes (Save Audio, etc.)
+                    # Many audio processing nodes don't support BFloat16
+                    return {
+                        "waveform": audio_tensor.cpu().float(),
+                        "sample_rate": 24000
+                    }
+                    
+                elif hasattr(output, 'sequences'):
+                    logger.error("VibeVoice returned only text tokens, no audio generated")
+                    raise Exception("VibeVoice failed to generate audio - only text tokens returned")
+                    
+                else:
+                    logger.error(f"Unexpected output format from VibeVoice: {type(output)}")
+                    raise Exception(f"VibeVoice returned unexpected output format: {type(output)}")
+                
+        except Exception as e:
+            # Re-raise interruption exceptions without wrapping
+            import comfy.model_management as mm
+            if isinstance(e, mm.InterruptProcessingException):
+                raise  # Let the interruption propagate
+            
+            # For real errors, log and re-raise with context
+            logger.error(f"VibeVoice generation failed: {e}")
+            raise Exception(f"VibeVoice generation failed: {str(e)}")

nodes/free_memory_node.py

@@ -0,0 +1,110 @@
+# Created by Fabio Sarracino
+# Node to free VibeVoice model memory
+
+import logging
+import torch
+import gc
+from typing import Any
+
+# Setup logging
+logger = logging.getLogger("VibeVoice")
+
+class VibeVoiceFreeMemoryNode:
+    """Node to explicitly free VibeVoice model memory"""
+    
+    # Class variables to store node instances
+    _single_speaker_instances = []
+    _multi_speaker_instances = []
+    
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "audio": ("AUDIO", {"tooltip": "Audio input that triggers memory cleanup and gets passed through"}),
+            }
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    RETURN_NAMES = ("audio",)
+    FUNCTION = "free_vibevoice_memory"
+    CATEGORY = "VibeVoiceWrapper"
+    DESCRIPTION = "Free all loaded VibeVoice models from memory when audio passes through"
+    
+    @classmethod
+    def register_single_speaker(cls, node_instance):
+        """Register a single speaker node instance"""
+        if node_instance not in cls._single_speaker_instances:
+            cls._single_speaker_instances.append(node_instance)
+    
+    @classmethod
+    def register_multi_speaker(cls, node_instance):
+        """Register a multi speaker node instance"""
+        if node_instance not in cls._multi_speaker_instances:
+            cls._multi_speaker_instances.append(node_instance)
+    
+    def free_vibevoice_memory(self, audio):
+        """Free memory from all VibeVoice nodes and pass through the audio"""
+        
+        try:
+            freed_count = 0
+            
+            # Try to access and free memory from globally cached instances
+            # ComfyUI might cache node instances
+            try:
+                import sys
+                from .base_vibevoice import BaseVibeVoiceNode
+                
+                # Search in all modules for BaseVibeVoiceNode instances
+                for module_name, module in sys.modules.items():
+                    if module and 'vibevoice' in module_name.lower():
+                        for attr_name in dir(module):
+                            if not attr_name.startswith('_'):
+                                try:
+                                    attr = getattr(module, attr_name)
+                                    if isinstance(attr, type) and issubclass(attr, BaseVibeVoiceNode):
+                                        # Check if the class has any cached instances
+                                        for instance_attr in dir(attr):
+                                            instance = getattr(attr, instance_attr)
+                                            if isinstance(instance, BaseVibeVoiceNode) and hasattr(instance, 'free_memory'):
+                                                instance.free_memory()
+                                                freed_count += 1
+                                except:
+                                    pass
+            except:
+                pass
+            
+            # Free from registered single speaker instances
+            for node in self._single_speaker_instances:
+                if hasattr(node, 'free_memory'):
+                    node.free_memory()
+                    freed_count += 1
+            
+            # Free from registered multi speaker instances  
+            for node in self._multi_speaker_instances:
+                if hasattr(node, 'free_memory'):
+                    node.free_memory()
+                    freed_count += 1
+            
+            # Force garbage collection
+            gc.collect()
+            
+            # Clear CUDA cache if available
+            if torch.cuda.is_available():
+                torch.cuda.empty_cache()
+                torch.cuda.synchronize()
+                logger.info(f"Freed VibeVoice memory from {freed_count} nodes and cleared CUDA cache")
+            else:
+                logger.info(f"Freed VibeVoice memory from {freed_count} nodes")
+            
+            # Pass through the audio unchanged
+            return (audio,)
+                
+        except Exception as e:
+            logger.error(f"Error freeing VibeVoice memory: {str(e)}")
+            # Still pass through audio even if error occurs
+            return (audio,)
+    
+    @classmethod
+    def IS_CHANGED(cls, **kwargs):
+        """Always execute this node"""
+        return float("nan")  # Forces re-execution every time

nodes/load_text_node.py

@@ -0,0 +1,173 @@
+# Created by Fabio Sarracino
+
+import os
+import logging
+import hashlib
+import folder_paths
+
+# Setup logging
+logger = logging.getLogger("VibeVoice")
+
+class LoadTextFromFileNode:
+    @classmethod
+    def INPUT_TYPES(cls):
+        # Get all text files from all directories
+        all_files = []
+        
+        # Add files from each directory with prefix
+        for dir_name in ["input", "output", "temp"]:
+            files = cls.get_files_for_directory(dir_name)
+            for f in files:
+                if f != "No text files found":
+                    all_files.append(f"{dir_name}/{f}")
+        
+        if not all_files:
+            all_files = ["No text files found in any directory"]
+        
+        return {
+            "required": {
+                "file": (sorted(all_files), {
+                    "tooltip": "Select a text file to load (format: directory/filename)"
+                }),
+            }
+        }
+    
+    @classmethod
+    def get_files_for_directory(cls, source_dir):
+        """Get list of text files for the selected directory"""
+        # Get the appropriate directory path
+        if source_dir == "input":
+            dir_path = folder_paths.get_input_directory()
+        elif source_dir == "output":
+            dir_path = folder_paths.get_output_directory()
+        elif source_dir == "temp":
+            dir_path = folder_paths.get_temp_directory()
+        else:
+            return []
+        
+        files = []
+        try:
+            for f in os.listdir(dir_path):
+                if os.path.isfile(os.path.join(dir_path, f)):
+                    # Check for text file extensions
+                    if f.lower().endswith(('.txt')):
+                        files.append(f)
+        except Exception as e:
+            logger.warning(f"Error listing files in {source_dir}: {e}")
+            
+        return files
+
+    RETURN_TYPES = ("STRING",)
+    RETURN_NAMES = ("text",)
+    FUNCTION = "load_text"
+    CATEGORY = "VibeVoiceWrapper"
+    DESCRIPTION = "Load text content from a .txt file"
+
+    def load_text(self, file: str):
+        """Load text content from file"""
+        
+        try:
+            # Check if no file selected
+            if not file or file == "No text files found in any directory":
+                raise Exception("Please select a valid text file.")
+            
+            # Parse directory and filename from the combined string
+            if "/" not in file:
+                raise Exception(f"Invalid file format: {file}")
+            
+            source_dir, filename = file.split("/", 1)
+            
+            # Get the appropriate directory path
+            if source_dir == "input":
+                dir_path = folder_paths.get_input_directory()
+            elif source_dir == "output":
+                dir_path = folder_paths.get_output_directory()
+            elif source_dir == "temp":
+                dir_path = folder_paths.get_temp_directory()
+            else:
+                raise Exception(f"Invalid source directory: {source_dir}")
+            
+            # Build full file path
+            file_path = os.path.join(dir_path, filename)
+            
+            if not os.path.exists(file_path):
+                raise Exception(f"File not found: {file_path}")
+            
+            # Read file with UTF-8 encoding (most common)
+            with open(file_path, 'r', encoding='utf-8') as f:
+                text_content = f.read()
+            
+            if not text_content.strip():
+                raise Exception("File is empty or contains only whitespace")
+            
+            return (text_content,)
+            
+        except UnicodeDecodeError as e:
+            raise Exception(f"Encoding error reading file: {str(e)}. File may not be UTF-8 encoded.")
+        except Exception as e:
+            logger.error(f"Failed to load text file: {str(e)}")
+            raise Exception(f"Error loading text file: {str(e)}")
+
+    @classmethod
+    def IS_CHANGED(cls, file):
+        """Cache key for ComfyUI"""
+        if not file or file == "No text files found in any directory":
+            return "no_file"
+        
+        # Parse directory and filename
+        if "/" not in file:
+            return f"{file}_invalid"
+        
+        source_dir, filename = file.split("/", 1)
+        
+        # Get the appropriate directory path
+        if source_dir == "input":
+            dir_path = folder_paths.get_input_directory()
+        elif source_dir == "output":
+            dir_path = folder_paths.get_output_directory()
+        elif source_dir == "temp":
+            dir_path = folder_paths.get_temp_directory()
+        else:
+            return f"{file}_invalid_dir"
+        
+        file_path = os.path.join(dir_path, filename)
+        
+        if not os.path.exists(file_path):
+            return f"{file}_not_found"
+        
+        # Use file hash for cache invalidation
+        try:
+            m = hashlib.sha256()
+            with open(file_path, 'rb') as f:
+                m.update(f.read())
+            return m.digest().hex()
+        except:
+            return f"{file}_error"
+    
+    @classmethod
+    def VALIDATE_INPUTS(cls, file, **kwargs):
+        """Validate that the file exists"""
+        if not file or file == "No text files found in any directory":
+            return "No valid text file selected"
+        
+        # Parse directory and filename
+        if "/" not in file:
+            return f"Invalid file format: {file}"
+        
+        source_dir, filename = file.split("/", 1)
+        
+        # Get the appropriate directory path
+        if source_dir == "input":
+            dir_path = folder_paths.get_input_directory()
+        elif source_dir == "output":
+            dir_path = folder_paths.get_output_directory()
+        elif source_dir == "temp":
+            dir_path = folder_paths.get_temp_directory()
+        else:
+            return f"Invalid source directory: {source_dir}"
+        
+        file_path = os.path.join(dir_path, filename)
+        if not os.path.exists(file_path):
+            return f"File not found: {filename} in {source_dir}"
+        
+        return True

nodes/multi_speaker_node.py

@@ -0,0 +1,342 @@
+# Created by Fabio Sarracino
+
+import logging
+import os
+import re
+import tempfile
+import torch
+import numpy as np
+from typing import List, Optional
+
+from .base_vibevoice import BaseVibeVoiceNode
+
+# Setup logging
+logger = logging.getLogger("VibeVoice")
+
+class VibeVoiceMultipleSpeakersNode(BaseVibeVoiceNode):
+    def __init__(self):
+        super().__init__()
+        # Register this instance for memory management
+        try:
+            from .free_memory_node import VibeVoiceFreeMemoryNode
+            VibeVoiceFreeMemoryNode.register_multi_speaker(self)
+        except:
+            pass
+    
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "text": ("STRING", {
+                    "multiline": True,
+                    "default": "[1]: Hello, this is the first speaker.\n[2]: Hi there, I'm the second speaker.\n[1]: Nice to meet you!\n[2]: Nice to meet you too!", 
+                    "tooltip": "Text with speaker labels. Use '[N]:' format where N is 1-4. Gets disabled when connected to another node.",
+                    "forceInput": False,
+                    "dynamicPrompts": True
+                }),
+                "model": (["VibeVoice-1.5B", "VibeVoice-Large", "VibeVoice-Large-Quant-4Bit"], {
+                    "default": "VibeVoice-Large",  # Large recommended for multi-speaker
+                    "tooltip": "Model to use. Large is recommended for multi-speaker generation, Quant-4Bit uses less VRAM (CUDA only)"
+                }),
+                "attention_type": (["auto", "eager", "sdpa", "flash_attention_2", "sage"], {
+                    "default": "auto",
+                    "tooltip": "Attention implementation. Auto selects the best available, eager is standard, sdpa is optimized PyTorch, flash_attention_2 requires compatible GPU, sage uses quantized attention for speedup (CUDA only)"
+                }),
+                "free_memory_after_generate": ("BOOLEAN", {"default": True, "tooltip": "Free model from memory after generation to save VRAM/RAM. Disable to keep model loaded for faster subsequent generations"}),
+                "diffusion_steps": ("INT", {"default": 20, "min": 5, "max": 100, "step": 1, "tooltip": "Number of denoising steps. More steps = better quality but slower. Default: 20"}),
+                "seed": ("INT", {"default": 42, "min": 0, "max": 2**32-1, "tooltip": "Random seed for generation. Default 42 is used in official examples"}),
+                "cfg_scale": ("FLOAT", {"default": 1.3, "min": 0.5, "max": 3.5, "step": 0.05, "tooltip": "Classifier-free guidance scale (official default: 1.3)"}),
+                "use_sampling": ("BOOLEAN", {"default": False, "tooltip": "Enable sampling mode. When False (default), uses deterministic generation like official examples"}),
+            },
+            "optional": {
+                "speaker1_voice": ("AUDIO", {"tooltip": "Optional: Voice sample for Speaker 1. If not provided, synthetic voice will be used."}),
+                "speaker2_voice": ("AUDIO", {"tooltip": "Optional: Voice sample for Speaker 2. If not provided, synthetic voice will be used."}),
+                "speaker3_voice": ("AUDIO", {"tooltip": "Optional: Voice sample for Speaker 3. If not provided, synthetic voice will be used."}),
+                "speaker4_voice": ("AUDIO", {"tooltip": "Optional: Voice sample for Speaker 4. If not provided, synthetic voice will be used."}),
+                "temperature": ("FLOAT", {"default": 0.95, "min": 0.1, "max": 2.0, "step": 0.05, "tooltip": "Only used when sampling is enabled"}),
+                "top_p": ("FLOAT", {"default": 0.95, "min": 0.1, "max": 1.0, "step": 0.05, "tooltip": "Only used when sampling is enabled"}),
+            }
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    RETURN_NAMES = ("audio",)
+    FUNCTION = "generate_speech"
+    CATEGORY = "VibeVoiceWrapper"
+    DESCRIPTION = "Generate multi-speaker conversations with up to 4 distinct voices using Microsoft VibeVoice"
+
+    def _prepare_voice_sample(self, voice_audio, speaker_idx: int) -> Optional[np.ndarray]:
+        """Prepare a single voice sample from input audio"""
+        return self._prepare_audio_from_comfyui(voice_audio)
+    
+    def generate_speech(self, text: str = "", model: str = "VibeVoice-7B-Preview",
+                       attention_type: str = "auto", free_memory_after_generate: bool = True,
+                       diffusion_steps: int = 20, seed: int = 42, cfg_scale: float = 1.3,
+                       use_sampling: bool = False, speaker1_voice=None, speaker2_voice=None, 
+                       speaker3_voice=None, speaker4_voice=None,
+                       temperature: float = 0.95, top_p: float = 0.95):
+        """Generate multi-speaker speech from text using VibeVoice"""
+        
+        try:
+            # Check text input
+            if not text or not text.strip():
+                raise Exception("No text provided. Please enter text with speaker labels (e.g., '[1]: Hello' or '[2]: Hi')")
+            
+            # First detect how many speakers are in the text
+            bracket_pattern = r'\[(\d+)\]\s*:'
+            speakers_numbers = sorted(list(set([int(m) for m in re.findall(bracket_pattern, text)])))
+            
+            # Limit to 1-4 speakers
+            if not speakers_numbers:
+                num_speakers = 1  # Default to 1 if no speaker format found
+            else:
+                num_speakers = min(max(speakers_numbers), 4)  # Max speaker number, capped at 4
+                if max(speakers_numbers) > 4:
+                    print(f"[VibeVoice] Warning: Found {max(speakers_numbers)} speakers, limiting to 4")
+            
+            # Direct conversion from [N]: to Speaker (N-1): for VibeVoice processor
+            # This avoids multiple conversion steps
+            converted_text = text
+            
+            # Find all [N]: patterns in the text
+            speakers_in_text = sorted(list(set([int(m) for m in re.findall(bracket_pattern, text)])))
+            
+            if not speakers_in_text:
+                # No [N]: format found, try Speaker N: format
+                speaker_pattern = r'Speaker\s+(\d+)\s*:'
+                speakers_in_text = sorted(list(set([int(m) for m in re.findall(speaker_pattern, text)])))
+                
+                if speakers_in_text:
+                    # Text already in Speaker N format, convert to 0-based
+                    for speaker_num in sorted(speakers_in_text, reverse=True):
+                        pattern = f'Speaker\\s+{speaker_num}\\s*:'
+                        replacement = f'Speaker {speaker_num - 1}:'
+                        converted_text = re.sub(pattern, replacement, converted_text)
+                else:
+                    # No speaker format found
+                    speakers_in_text = [1]
+                    
+                    # Parse pause keywords even for single speaker
+                    pause_segments = self._parse_pause_keywords(text)
+                    
+                    # Store speaker segments for pause processing
+                    speaker_segments_with_pauses = []
+                    segments = []
+                    
+                    for seg_type, seg_content in pause_segments:
+                        if seg_type == 'pause':
+                            speaker_segments_with_pauses.append(('pause', seg_content, None))
+                        else:
+                            # Clean up newlines
+                            text_clean = seg_content.replace('\n', ' ').replace('\r', ' ')
+                            text_clean = ' '.join(text_clean.split())
+                            
+                            if text_clean:
+                                speaker_segments_with_pauses.append(('text', text_clean, 1))
+                                segments.append(f"Speaker 0: {text_clean}")
+                    
+                    # Join all segments for fallback
+                    converted_text = '\n'.join(segments) if segments else f"Speaker 0: {text}"
+            else:
+                # Convert [N]: directly to Speaker (N-1): and handle multi-line text
+                # Split text to preserve speaker segments while cleaning up newlines within each segment
+                segments = []
+                
+                # Find all speaker markers with their positions
+                speaker_matches = list(re.finditer(f'\\[({"|".join(map(str, speakers_in_text))})\\]\\s*:', converted_text))
+                
+                # Store speaker segments for pause processing
+                speaker_segments_with_pauses = []
+                
+                for i, match in enumerate(speaker_matches):
+                    speaker_num = int(match.group(1))
+                    start = match.end()
+                    
+                    # Find where this speaker's text ends (at next speaker or end of text)
+                    if i + 1 < len(speaker_matches):
+                        end = speaker_matches[i + 1].start()
+                    else:
+                        end = len(converted_text)
+                    
+                    # Extract the speaker's text (keep pause keywords for now)
+                    speaker_text = converted_text[start:end].strip()
+                    
+                    # Parse pause keywords within this speaker's text
+                    pause_segments = self._parse_pause_keywords(speaker_text)
+                    
+                    # Process each segment (text or pause) for this speaker
+                    for seg_type, seg_content in pause_segments:
+                        if seg_type == 'pause':
+                            # Add pause segment
+                            speaker_segments_with_pauses.append(('pause', seg_content, None))
+                        else:
+                            # Clean up the text segment
+                            text_clean = seg_content.replace('\n', ' ').replace('\r', ' ')
+                            text_clean = ' '.join(text_clean.split())
+                            
+                            if text_clean:  # Only add non-empty text
+                                # Add text segment with speaker info
+                                speaker_segments_with_pauses.append(('text', text_clean, speaker_num))
+                                # Also build the traditional segments for fallback
+                                segments.append(f'Speaker {speaker_num - 1}: {text_clean}')
+                
+                # Join all segments with newlines (required for multi-speaker format) - for fallback
+                converted_text = '\n'.join(segments) if segments else ""
+            
+            # Build speaker names list - these are just for logging, not used by processor
+            # The processor uses the speaker labels in the text itself
+            speakers = [f"Speaker {i}" for i in range(len(speakers_in_text))]
+            
+            # Get model mapping and load model with attention type
+            model_mapping = self._get_model_mapping()
+            model_path = model_mapping.get(model, model)
+            self.load_model(model, model_path, attention_type)
+            
+            voice_inputs = [speaker1_voice, speaker2_voice, speaker3_voice, speaker4_voice]
+            
+            # Prepare voice samples in order of appearance
+            voice_samples = []
+            for i, speaker_num in enumerate(speakers_in_text):
+                idx = speaker_num - 1  # Convert to 0-based for voice array
+                
+                # Try to use provided voice sample
+                if idx < len(voice_inputs) and voice_inputs[idx] is not None:
+                    voice_sample = self._prepare_voice_sample(voice_inputs[idx], idx)
+                    if voice_sample is None:
+                        # Use the actual speaker index for consistent synthetic voice
+                        voice_sample = self._create_synthetic_voice_sample(idx)
+                else:
+                    # Use the actual speaker index for consistent synthetic voice
+                    voice_sample = self._create_synthetic_voice_sample(idx)
+                    
+                voice_samples.append(voice_sample)
+            
+            # Ensure voice_samples count matches detected speakers
+            if len(voice_samples) != len(speakers_in_text):
+                logger.error(f"Mismatch: {len(speakers_in_text)} speakers but {len(voice_samples)} voice samples!")
+                raise Exception(f"Voice sample count mismatch: expected {len(speakers_in_text)}, got {len(voice_samples)}")
+            
+            # Check if we have pause segments to process
+            if 'speaker_segments_with_pauses' in locals() and speaker_segments_with_pauses:
+                # Process segments with pauses
+                all_audio_segments = []
+                sample_rate = 24000  # VibeVoice uses 24kHz
+                
+                # Group consecutive text segments from same speaker for efficiency
+                grouped_segments = []
+                current_group = []
+                current_speaker = None
+                
+                for seg_type, seg_content, speaker_num in speaker_segments_with_pauses:
+                    if seg_type == 'pause':
+                        # Save current group if any
+                        if current_group:
+                            grouped_segments.append(('text_group', current_group, current_speaker))
+                            current_group = []
+                            current_speaker = None
+                        # Add pause
+                        grouped_segments.append(('pause', seg_content, None))
+                    else:
+                        # Text segment
+                        if speaker_num == current_speaker:
+                            # Same speaker, add to current group
+                            current_group.append(seg_content)
+                        else:
+                            # Different speaker, save current group and start new one
+                            if current_group:
+                                grouped_segments.append(('text_group', current_group, current_speaker))
+                            current_group = [seg_content]
+                            current_speaker = speaker_num
+                
+                # Save last group if any
+                if current_group:
+                    grouped_segments.append(('text_group', current_group, current_speaker))
+                
+                # Process grouped segments
+                for seg_type, seg_content, speaker_num in grouped_segments:
+                    if seg_type == 'pause':
+                        # Generate silence
+                        duration_ms = seg_content
+                        logger.info(f"Adding {duration_ms}ms pause")
+                        silence_audio = self._generate_silence(duration_ms, sample_rate)
+                        all_audio_segments.append(silence_audio)
+                    else:
+                        # Process text group for a speaker
+                        combined_text = ' '.join(seg_content)
+                        formatted_text = f"Speaker {speaker_num - 1}: {combined_text}"
+                        
+                        # Get voice sample for this speaker
+                        speaker_idx = speakers_in_text.index(speaker_num)
+                        speaker_voice_samples = [voice_samples[speaker_idx]]
+                        
+                        logger.info(f"Generating audio for Speaker {speaker_num}: {len(combined_text.split())} words")
+                        
+                        # Generate audio for this speaker's text
+                        segment_audio = self._generate_with_vibevoice(
+                            formatted_text, speaker_voice_samples, cfg_scale, seed,
+                            diffusion_steps, use_sampling, temperature, top_p
+                        )
+                        
+                        all_audio_segments.append(segment_audio)
+                
+                # Concatenate all audio segments
+                if all_audio_segments:
+                    logger.info(f"Concatenating {len(all_audio_segments)} audio segments (including pauses)...")
+                    
+                    # Extract waveforms
+                    waveforms = []
+                    for audio_segment in all_audio_segments:
+                        if isinstance(audio_segment, dict) and "waveform" in audio_segment:
+                            waveforms.append(audio_segment["waveform"])
+                    
+                    if waveforms:
+                        # Filter out None values if any
+                        valid_waveforms = [w for w in waveforms if w is not None]
+                        
+                        if valid_waveforms:
+                            # Concatenate along time dimension
+                            combined_waveform = torch.cat(valid_waveforms, dim=-1)
+                            
+                            audio_dict = {
+                                "waveform": combined_waveform,
+                                "sample_rate": sample_rate
+                            }
+                            logger.info(f"Successfully generated multi-speaker audio with pauses")
+                        else:
+                            raise Exception("No valid audio waveforms generated")
+                    else:
+                        raise Exception("Failed to extract waveforms from audio segments")
+                else:
+                    raise Exception("No audio segments generated")
+            else:
+                # Fallback to original method without pause support
+                logger.info("Processing without pause support (no pause keywords found)")
+                audio_dict = self._generate_with_vibevoice(
+                    converted_text, voice_samples, cfg_scale, seed, diffusion_steps,
+                    use_sampling, temperature, top_p
+                )
+            
+            # Free memory if requested
+            if free_memory_after_generate:
+                self.free_memory()
+            
+            return (audio_dict,)
+                    
+        except Exception as e:
+            # Check if this is an interruption by the user
+            import comfy.model_management as mm
+            if isinstance(e, mm.InterruptProcessingException):
+                # User interrupted - just log it and re-raise to stop the workflow
+                logger.info("Generation interrupted by user")
+                raise  # Propagate the interruption to stop the workflow
+            else:
+                # Real error - show it
+                logger.error(f"Multi-speaker speech generation failed: {str(e)}")
+                raise Exception(f"Error generating multi-speaker speech: {str(e)}")
+
+    @classmethod
+    def IS_CHANGED(cls, text="", model="VibeVoice-7B-Preview",
+                   speaker1_voice=None, speaker2_voice=None, 
+                   speaker3_voice=None, speaker4_voice=None, **kwargs):
+        """Cache key for ComfyUI"""
+        voices_hash = hash(str([speaker1_voice, speaker2_voice, speaker3_voice, speaker4_voice]))
+        return f"{hash(text)}_{model}_{voices_hash}_{kwargs.get('cfg_scale', 1.3)}_{kwargs.get('seed', 0)}"

nodes/single_speaker_node.py

@@ -0,0 +1,220 @@
+# Created by Fabio Sarracino
+
+import logging
+import os
+import tempfile
+import torch
+import numpy as np
+import re
+from typing import List, Optional
+
+from .base_vibevoice import BaseVibeVoiceNode
+
+# Setup logging
+logger = logging.getLogger("VibeVoice")
+
+class VibeVoiceSingleSpeakerNode(BaseVibeVoiceNode):
+    def __init__(self):
+        super().__init__()
+        # Register this instance for memory management
+        try:
+            from .free_memory_node import VibeVoiceFreeMemoryNode
+            VibeVoiceFreeMemoryNode.register_single_speaker(self)
+        except:
+            pass
+    
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "text": ("STRING", {
+                    "multiline": True,
+                    "default": "Hello, this is a test of the VibeVoice text-to-speech system.", 
+                    "tooltip": "Text to convert to speech. Gets disabled when connected to another node.",
+                    "forceInput": False,
+                    "dynamicPrompts": True
+                }),
+                "model": (["VibeVoice-1.5B", "VibeVoice-Large", "VibeVoice-Large-Quant-4Bit"], {
+                    "default": "VibeVoice-1.5B", 
+                    "tooltip": "Model to use. 1.5B is faster, Large has better quality, Quant-4Bit uses less VRAM (CUDA only)"
+                }),
+                "attention_type": (["auto", "eager", "sdpa", "flash_attention_2", "sage"], {
+                    "default": "auto",
+                    "tooltip": "Attention implementation. Auto selects the best available, eager is standard, sdpa is optimized PyTorch, flash_attention_2 requires compatible GPU, sage uses quantized attention for speedup (CUDA only)"
+                }),
+                "free_memory_after_generate": ("BOOLEAN", {"default": True, "tooltip": "Free model from memory after generation to save VRAM/RAM. Disable to keep model loaded for faster subsequent generations"}),
+                "diffusion_steps": ("INT", {"default": 20, "min": 5, "max": 100, "step": 1, "tooltip": "Number of denoising steps. More steps = better quality but slower. Default: 20"}),
+                "seed": ("INT", {"default": 42, "min": 0, "max": 2**32-1, "tooltip": "Random seed for generation. Default 42 is used in official examples"}),
+                "cfg_scale": ("FLOAT", {"default": 1.3, "min": 0.5, "max": 3.5, "step": 0.05, "tooltip": "Classifier-free guidance scale (official default: 1.3)"}),
+                "use_sampling": ("BOOLEAN", {"default": False, "tooltip": "Enable sampling mode. When False (default), uses deterministic generation like official examples"}),
+            },
+            "optional": {
+                "voice_to_clone": ("AUDIO", {"tooltip": "Optional: Reference voice to clone. If not provided, synthetic voice will be used."}),
+                "temperature": ("FLOAT", {"default": 0.95, "min": 0.1, "max": 2.0, "step": 0.05, "tooltip": "Only used when sampling is enabled"}),
+                "top_p": ("FLOAT", {"default": 0.95, "min": 0.1, "max": 1.0, "step": 0.05, "tooltip": "Only used when sampling is enabled"}),
+                "max_words_per_chunk": ("INT", {"default": 250, "min": 100, "max": 500, "step": 50, "tooltip": "Maximum words per chunk for long texts. Lower values prevent speed issues but create more chunks."}),
+            }
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    RETURN_NAMES = ("audio",)
+    FUNCTION = "generate_speech"
+    CATEGORY = "VibeVoiceWrapper"
+    DESCRIPTION = "Generate speech from text using Microsoft VibeVoice with optional voice cloning"
+
+    def _prepare_voice_samples(self, speakers: list, voice_to_clone) -> List[np.ndarray]:
+        """Prepare voice samples from input audio or create synthetic ones"""
+        
+        if voice_to_clone is not None:
+            # Use the base class method to prepare audio
+            audio_np = self._prepare_audio_from_comfyui(voice_to_clone)
+            if audio_np is not None:
+                return [audio_np]
+        
+        # Create synthetic voice samples for speakers
+        voice_samples = []
+        for i, speaker in enumerate(speakers):
+            voice_sample = self._create_synthetic_voice_sample(i)
+            voice_samples.append(voice_sample)
+            
+        return voice_samples
+    
+    def generate_speech(self, text: str = "", model: str = "VibeVoice-1.5B", 
+                       attention_type: str = "auto", free_memory_after_generate: bool = True,
+                       diffusion_steps: int = 20, seed: int = 42, cfg_scale: float = 1.3,
+                       use_sampling: bool = False, voice_to_clone=None,
+                       temperature: float = 0.95, top_p: float = 0.95,
+                       max_words_per_chunk: int = 250):
+        """Generate speech from text using VibeVoice"""
+        
+        try:
+            # Use text directly (it now serves as both manual input and connection input)
+            if text and text.strip():
+                final_text = text
+            else:
+                raise Exception("No text provided. Please enter text or connect from LoadTextFromFile node.")
+            
+            # Get model mapping and load model with attention type
+            model_mapping = self._get_model_mapping()
+            model_path = model_mapping.get(model, model)
+            self.load_model(model, model_path, attention_type)
+            
+            # For single speaker, we just use ["Speaker 1"]
+            speakers = ["Speaker 1"]
+            
+            # Parse pause keywords from text
+            segments = self._parse_pause_keywords(final_text)
+            
+            # Process segments
+            all_audio_segments = []
+            voice_samples = None  # Will be created on first text segment
+            sample_rate = 24000  # VibeVoice uses 24kHz
+            
+            for seg_idx, (seg_type, seg_content) in enumerate(segments):
+                if seg_type == 'pause':
+                    # Generate silence for pause
+                    duration_ms = seg_content
+                    logger.info(f"Adding {duration_ms}ms pause")
+                    silence_audio = self._generate_silence(duration_ms, sample_rate)
+                    all_audio_segments.append(silence_audio)
+                    
+                elif seg_type == 'text':
+                    # Process text segment (with chunking if needed)
+                    word_count = len(seg_content.split())
+                    
+                    if word_count > max_words_per_chunk:
+                        # Split long text into chunks
+                        logger.info(f"Text segment {seg_idx+1} has {word_count} words, splitting into chunks...")
+                        text_chunks = self._split_text_into_chunks(seg_content, max_words_per_chunk)
+                        
+                        for chunk_idx, chunk in enumerate(text_chunks):
+                            logger.info(f"Processing chunk {chunk_idx+1}/{len(text_chunks)} of segment {seg_idx+1}...")
+                            
+                            # Format chunk for VibeVoice
+                            formatted_text = self._format_text_for_vibevoice(chunk, speakers)
+                            
+                            # Create voice samples on first text segment
+                            if voice_samples is None:
+                                voice_samples = self._prepare_voice_samples(speakers, voice_to_clone)
+                            
+                            # Generate audio for this chunk
+                            chunk_audio = self._generate_with_vibevoice(
+                                formatted_text, voice_samples, cfg_scale, 
+                                seed,  # Use same seed for voice consistency
+                                diffusion_steps, use_sampling, temperature, top_p
+                            )
+                            
+                            all_audio_segments.append(chunk_audio)
+                    else:
+                        # Process as single chunk
+                        logger.info(f"Processing text segment {seg_idx+1} ({word_count} words)")
+                        
+                        # Format text for VibeVoice
+                        formatted_text = self._format_text_for_vibevoice(seg_content, speakers)
+                        
+                        # Create voice samples on first text segment
+                        if voice_samples is None:
+                            voice_samples = self._prepare_voice_samples(speakers, voice_to_clone)
+                        
+                        # Generate audio
+                        segment_audio = self._generate_with_vibevoice(
+                            formatted_text, voice_samples, cfg_scale, seed, diffusion_steps, 
+                            use_sampling, temperature, top_p
+                        )
+                        
+                        all_audio_segments.append(segment_audio)
+            
+            # Concatenate all audio segments (including pauses)
+            if all_audio_segments:
+                logger.info(f"Concatenating {len(all_audio_segments)} audio segments (including pauses)...")
+                
+                # Extract waveforms from all segments
+                waveforms = []
+                for audio_segment in all_audio_segments:
+                    if isinstance(audio_segment, dict) and "waveform" in audio_segment:
+                        waveforms.append(audio_segment["waveform"])
+                
+                if waveforms:
+                    # Filter out None values if any
+                    valid_waveforms = [w for w in waveforms if w is not None]
+                    
+                    if valid_waveforms:
+                        # Concatenate along the time dimension (last dimension)
+                        combined_waveform = torch.cat(valid_waveforms, dim=-1)
+                        
+                        # Create final audio dict
+                        audio_dict = {
+                            "waveform": combined_waveform,
+                            "sample_rate": sample_rate
+                        }
+                        logger.info(f"Successfully generated audio with {len(segments)} segments")
+                    else:
+                        raise Exception("No valid audio waveforms generated")
+                else:
+                    raise Exception("Failed to extract waveforms from audio segments")
+            else:
+                raise Exception("No audio segments generated")
+            
+            # Free memory if requested
+            if free_memory_after_generate:
+                self.free_memory()
+            
+            return (audio_dict,)
+                    
+        except Exception as e:
+            # Check if this is an interruption by the user
+            import comfy.model_management as mm
+            if isinstance(e, mm.InterruptProcessingException):
+                # User interrupted - just log it and re-raise to stop the workflow
+                logger.info("Generation interrupted by user")
+                raise  # Propagate the interruption to stop the workflow
+            else:
+                # Real error - show it
+                logger.error(f"Single speaker speech generation failed: {str(e)}")
+                raise Exception(f"Error generating speech: {str(e)}")
+
+    @classmethod
+    def IS_CHANGED(cls, text="", model="VibeVoice-1.5B", voice_to_clone=None, **kwargs):
+        """Cache key for ComfyUI"""
+        voice_hash = hash(str(voice_to_clone)) if voice_to_clone else 0
+        return f"{hash(text)}_{model}_{voice_hash}_{kwargs.get('cfg_scale', 1.3)}_{kwargs.get('seed', 0)}"

pyproject.toml

@@ -0,0 +1,18 @@
+[project]
+name = "VibeVoice-ComfyUI"
+version = "1.3.0"
+description = "ComfyUI wrapper for Microsoft VibeVoice TTS model. Supports single speaker, multi-speaker, and text file loading"
+license = {file = "LICENSE"} 
+authors = [{name = "Fabio Sarracino"}]
+dependencies = ["accelerate==1.6.0", "torch>=2.0.0", "torchaudio>=2.0.0", "numpy>=1.20.0", "transformers>=4.51.3", "librosa>=0.9.0", "soundfile>=0.12.0", "av>=14.3.0", "peft>=0.17.0", "huggingface_hub>=0.25.1", "diffusers", "tqdm", "scipy", "ml-collections", "absl-py", "aiortc", "bitsandbytes"]
+
+[project.urls]
+Repository = "https://github.com/Enemyx-net/VibeVoice-ComfyUI"
+"Bug Tracker" = "https://github.com/Enemyx-net/VibeVoice-ComfyUI/issues"
+
+[tool.comfy]
+PublisherId = "enemyx"
+DisplayName = "VibeVoice ComfyUI"
+Icon = ""
+includes = []
+

requirements.txt

@@ -0,0 +1,17 @@
+accelerate>=1.6.0
+transformers>=4.51.3
+diffusers
+tqdm
+scipy
+ml-collections
+torch>=2.0.0
+torchaudio>=2.0.0
+numpy>=1.20.0
+librosa>=0.9.0
+soundfile>=0.12.0
+av>=14.3.0
+peft>=0.17.0
+huggingface_hub>=0.25.1
+absl-py
+aiortc
+bitsandbytes

vvembed/LICENSE

@@ -0,0 +1,26 @@
+MIT License
+
+Copyright (c) Microsoft Corporation.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+---
+
+This is the original VibeVoice code from Microsoft, embedded here as the
+repository has been removed from GitHub. The code is used under the MIT license.

vvembed/README.md

@@ -0,0 +1,21 @@
+# Embedded VibeVoice
+
+This folder contains the embedded VibeVoice code from Microsoft.
+
+## Why Embedded?
+
+The original VibeVoice repository (https://github.com/microsoft/VibeVoice) has been removed from GitHub. Since VibeVoice is licensed under MIT, we have embedded the code here to ensure continued functionality of the ComfyUI wrapper.
+
+## License
+
+The code in this folder is licensed under the MIT License (see LICENSE file). Original copyright belongs to Microsoft Corporation.
+
+## Modifications
+
+The only modifications made to the original code are:
+- Changed absolute imports from `vibevoice` to relative imports
+- No functional changes to the core logic
+
+## Note
+
+This is a preservation copy to ensure the continued availability of VibeVoice for the ComfyUI community.

vvembed/modular/configuration_vibevoice.py

@@ -0,0 +1,270 @@
+# Original code by Microsoft
+# updated by Fabio Sarracino - Enemyx-net
+
+""" VibeVoice_AcousticTokenizer model configuration"""
+
+from typing import Dict, List, Optional, Tuple
+
+from transformers.configuration_utils import PretrainedConfig 
+from transformers.utils import logging
+
+from transformers.models.qwen2.configuration_qwen2 import Qwen2Config
+
+logger = logging.get_logger(__name__)
+
+# to be improved...
+
+
+class VibeVoiceAcousticTokenizerConfig(PretrainedConfig):
+    model_type = "vibevoice_acoustic_tokenizer"
+
+    def __init__(
+        self,
+        channels: int = 1,
+        corpus_normalize: float = 0.0,
+        causal: bool = True,
+        vae_dim: int = 64,
+        fix_std: float = 0.5,
+        std_dist_type: str = 'gaussian',
+        # common 
+        mixer_layer: str = 'depthwise_conv',
+        conv_norm: str = 'none',
+        pad_mode: str = 'constant',
+        disable_last_norm: bool = True,
+        layernorm: str = 'RMSNorm',
+        layernorm_eps: float = 1e-5,
+        layernorm_elementwise_affine: bool = True,
+        conv_bias: bool = True,
+        layer_scale_init_value: float = 1e-6,
+        weight_init_value: float = 1e-2,
+        # encoder specific
+        encoder_n_filters: int = 32,
+        encoder_ratios: Optional[List[int]] = [8,5,5,4,2,2],
+        encoder_depths: str = "3-3-3-3-3-3-8",
+        # decoder specific
+        decoder_n_filters: int = 32,
+        decoder_ratios: Optional[List[int]] = None, # if None, same as encoder
+        decoder_depths: Optional[str] = None,
+        **kwargs
+    ):
+        super().__init__(**kwargs)
+        self.channels = channels
+        self.corpus_normalize = corpus_normalize
+        self.causal = causal
+        self.vae_dim = vae_dim
+        self.fix_std = fix_std
+        self.std_dist_type = std_dist_type
+        
+        # common parameters
+        self.conv_norm = conv_norm
+        self.pad_mode = pad_mode
+        self.layernorm_eps = layernorm_eps
+        self.disable_last_norm = disable_last_norm
+        self.layernorm = layernorm
+        self.layernorm_elementwise_affine = layernorm_elementwise_affine
+        self.conv_bias = conv_bias
+        self.layer_scale_init_value = layer_scale_init_value
+        self.weight_init_value = weight_init_value
+        self.mixer_layer = mixer_layer
+
+        # encoder specific parameters
+        self.encoder_n_filters = encoder_n_filters
+        self.encoder_ratios = encoder_ratios
+        self.encoder_depths = encoder_depths
+        
+        # decoder specific parameters
+        self.decoder_ratios = decoder_ratios if decoder_ratios is not None else encoder_ratios
+        self.decoder_n_filters = decoder_n_filters
+        self.decoder_depths = decoder_depths
+
+
+class VibeVoiceSemanticTokenizerConfig(PretrainedConfig):
+    model_type = "vibevoice_semantic_tokenizer"
+    
+    def __init__(
+        self,
+        channels: int = 1,
+        corpus_normalize: float = 0.0,
+        causal: bool = True,
+        vae_dim: int = 64,
+        fix_std: float = 0,
+        std_dist_type: str = 'none',
+        # common 
+        mixer_layer: str = 'depthwise_conv',
+        conv_norm: str = 'none',
+        pad_mode: str = 'constant',
+        disable_last_norm: bool = True,
+        layernorm: str = 'RMSNorm',
+        layernorm_eps: float = 1e-5,
+        layernorm_elementwise_affine: bool = True,
+        conv_bias: bool = True,
+        layer_scale_init_value: float = 1e-6,
+        weight_init_value: float = 1e-2,
+        # encoder specific
+        encoder_n_filters: int = 32,
+        encoder_ratios: Optional[List[int]] = [8,5,5,4,2,2],
+        encoder_depths: str = "3-3-3-3-3-3-8",
+        **kwargs
+    ):
+        super().__init__(**kwargs)
+        self.channels = channels
+        self.corpus_normalize = corpus_normalize
+        self.causal = causal
+        self.vae_dim = vae_dim
+        self.fix_std = fix_std
+        self.std_dist_type = std_dist_type
+        
+        # common parameters
+        self.conv_norm = conv_norm
+        self.pad_mode = pad_mode
+        self.layernorm_eps = layernorm_eps
+        self.disable_last_norm = disable_last_norm
+        self.layernorm = layernorm
+        self.layernorm_elementwise_affine = layernorm_elementwise_affine
+        self.conv_bias = conv_bias
+        self.layer_scale_init_value = layer_scale_init_value
+        self.weight_init_value = weight_init_value
+        self.mixer_layer = mixer_layer
+
+        # encoder specific parameters
+        self.encoder_n_filters = encoder_n_filters
+        self.encoder_ratios = encoder_ratios
+        self.encoder_depths = encoder_depths
+        
+
+class VibeVoiceDiffusionHeadConfig(PretrainedConfig):
+    model_type = "vibevoice_diffusion_head"
+
+    def __init__(
+        self,
+        hidden_size=768,
+        head_layers=4,
+        head_ffn_ratio=3.0,
+        rms_norm_eps=1e-5,
+        latent_size=64,
+        speech_vae_dim=None,
+        prediction_type="v_prediction",
+        diffusion_type="ddpm",
+        ddpm_num_steps=1000,
+        ddpm_num_inference_steps=20,
+        ddpm_beta_schedule="cosine",
+        ddpm_batch_mul=4,
+        **kwargs
+    ):
+        self.hidden_size = hidden_size
+        self.head_layers = head_layers
+        self.head_ffn_ratio = head_ffn_ratio
+        self.rms_norm_eps = rms_norm_eps
+        self.latent_size = latent_size
+        self.speech_vae_dim = speech_vae_dim
+        self.prediction_type = prediction_type
+        self.diffusion_type = diffusion_type
+        self.ddpm_num_steps = ddpm_num_steps
+        self.ddpm_num_inference_steps = ddpm_num_inference_steps
+        self.ddpm_beta_schedule = ddpm_beta_schedule
+        self.ddpm_batch_mul = ddpm_batch_mul
+        
+        super().__init__(**kwargs)
+
+class VibeVoiceConfig(PretrainedConfig):
+    model_type = "vibevoice"
+    is_composition = True
+    sub_configs = {
+        "acoustic_tokenizer_config": VibeVoiceAcousticTokenizerConfig, 
+        "semantic_tokenizer_config": VibeVoiceSemanticTokenizerConfig,
+        "decoder_config": Qwen2Config,
+        "diffusion_head_config": VibeVoiceDiffusionHeadConfig,
+    }
+    # keys_to_ignore_at_inference = ["past_key_values"]
+    # Default tensor parallel plan for base model `Qwen2`
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    
+    def __init__(
+        self,
+        acoustic_tokenizer_config=None,
+        semantic_tokenizer_config=None,
+        decoder_config=None,
+        diffusion_head_config=None,
+        **kwargs
+    ):
+
+        # kwargs["_attn_implementation"] = "flash_attention_2"
+        kwargs["_attn_implementation_autoset"] = False 
+
+        if acoustic_tokenizer_config is None:
+            self.acoustic_tokenizer_config = self.sub_configs["acoustic_tokenizer_config"]()
+        elif isinstance(acoustic_tokenizer_config, dict):
+            acoustic_tokenizer_config["model_type"] = "vibevoice_acoustic_tokenizer"
+            self.acoustic_tokenizer_config = self.sub_configs["acoustic_tokenizer_config"](**acoustic_tokenizer_config)
+        elif isinstance(acoustic_tokenizer_config, VibeVoiceAcousticTokenizerConfig):
+            # If an instance of the config class is provided
+            self.acoustic_tokenizer_config = acoustic_tokenizer_config
+
+        if semantic_tokenizer_config is None:
+            self.semantic_tokenizer_config = self.sub_configs["semantic_tokenizer_config"]()
+        elif isinstance(semantic_tokenizer_config, dict):
+            semantic_tokenizer_config["model_type"] = "vibevoice_semantic_tokenizer"
+            self.semantic_tokenizer_config = self.sub_configs["semantic_tokenizer_config"](**semantic_tokenizer_config)
+        elif isinstance(semantic_tokenizer_config, VibeVoiceSemanticTokenizerConfig):
+            # If an instance of the config class is provided
+            self.semantic_tokenizer_config = semantic_tokenizer_config
+
+        if decoder_config is None:
+            self.decoder_config = self.sub_configs["decoder_config"]()
+        elif isinstance(decoder_config, dict):
+            # If a dictionary is provided, instantiate the config class with it
+            # self.decoder_config = self.sub_configs["decoder_config"](**decoder_config)
+            if decoder_config.get("model_type", '') == "qwen2":
+                self.decoder_config = Qwen2Config(**decoder_config)
+            else:
+                raise ValueError(f"Unsupported decoder model type: {decoder_config.get('model_type', '')}")
+        elif isinstance(decoder_config, (Qwen2Config,)):
+            # If an instance of the config class is provided
+            self.decoder_config = decoder_config
+
+        if diffusion_head_config is None:
+            self.diffusion_head_config = self.sub_configs["diffusion_head_config"]()
+        elif isinstance(diffusion_head_config, dict):
+            diffusion_head_config["model_type"] = "vibevoice_diffusion_head"
+            self.diffusion_head_config = self.sub_configs["diffusion_head_config"](**diffusion_head_config)
+        elif isinstance(diffusion_head_config, VibeVoiceDiffusionHeadConfig):
+            # If an instance of the config class is provided
+            self.diffusion_head_config = diffusion_head_config
+
+        # other parameters
+        self.acoustic_vae_dim = getattr(self.acoustic_tokenizer_config, 'vae_dim', 64)
+        self.semantic_vae_dim = getattr(self.semantic_tokenizer_config, 'vae_dim', 128)
+        
+        # Add attributes required by newer transformers versions from decoder_config
+        # These are used by GenerationMixin in newer versions
+        if hasattr(self.decoder_config, 'num_hidden_layers'):
+            self.num_hidden_layers = self.decoder_config.num_hidden_layers
+        if hasattr(self.decoder_config, 'vocab_size'):
+            self.vocab_size = self.decoder_config.vocab_size
+        if hasattr(self.decoder_config, 'hidden_size'):
+            self.hidden_size = self.decoder_config.hidden_size
+        if hasattr(self.decoder_config, 'num_attention_heads'):
+            self.num_attention_heads = self.decoder_config.num_attention_heads
+        if hasattr(self.decoder_config, 'num_key_value_heads'):
+            self.num_key_value_heads = self.decoder_config.num_key_value_heads
+        if hasattr(self.decoder_config, 'intermediate_size'):
+            self.intermediate_size = self.decoder_config.intermediate_size
+        if hasattr(self.decoder_config, 'max_position_embeddings'):
+            self.max_position_embeddings = self.decoder_config.max_position_embeddings
+
+        super().__init__(**kwargs)
+
+__all__ = [
+    "VibeVoiceAcousticTokenizerConfig", 
+    "VibeVoiceSemanticTokenizerConfig", 
+    "VibeVoiceDiffusionHeadConfig", 
+    "VibeVoiceConfig"
+]

vvembed/modular/modeling_vibevoice.py

@@ -0,0 +1,488 @@
+from dataclasses import dataclass
+from typing import Dict, List, Optional, Tuple, Union, Callable
+from tqdm import tqdm
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.distributed as dist
+
+from transformers.models.auto import AutoModel, AutoModelForCausalLM
+
+from transformers.activations import ACT2FN
+from transformers.modeling_outputs import CausalLMOutput, BaseModelOutputWithPast, ModelOutput
+from transformers.models.llama.modeling_llama import LlamaRMSNorm
+from transformers import modeling_utils
+from transformers.modeling_utils import PreTrainedModel
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.utils import logging
+
+
+from .modular_vibevoice_tokenizer import VibeVoiceTokenizerStreamingCache, VibeVoiceAcousticTokenizerModel, VibeVoiceSemanticTokenizerModel
+from .modular_vibevoice_diffusion_head import VibeVoiceDiffusionHead
+from schedule.dpm_solver import DPMSolverMultistepScheduler
+
+from .configuration_vibevoice import VibeVoiceConfig
+
+
+logger = logging.get_logger(__name__)
+
+if not hasattr(modeling_utils, "ALL_PARALLEL_STYLES") or modeling_utils.ALL_PARALLEL_STYLES is None:
+    modeling_utils.ALL_PARALLEL_STYLES = ["tp", "none", "colwise", "rowwise"]
+
+@dataclass
+class VibeVoiceCausalLMOutputWithPast(ModelOutput):
+    loss: Optional[torch.FloatTensor] = None
+    diffusion_loss: Optional[torch.FloatTensor] = None
+    speech_token_num: Optional[int] = None
+    logits: torch.FloatTensor = None
+    past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None
+    hidden_states: Optional[Tuple[torch.FloatTensor, ...]] = None
+    attentions: Optional[Tuple[torch.FloatTensor, ...]] = None
+
+
+@dataclass
+class VibeVoiceGenerationOutput(ModelOutput):
+    """
+    Output type for VibeVoice generation.
+    
+    Args:
+        sequences (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            The generated sequences. 
+        speech_outputs (`List[torch.FloatTensor]`, *optional*):
+            List of generated speech waveforms or latents for each speech segment.
+    """
+    sequences: torch.LongTensor = None
+    speech_outputs: Optional[List[torch.FloatTensor]] = None
+
+
+class SpeechConnector(nn.Module):
+    def __init__(self, input_dim, output_dim):
+        super().__init__()
+        self.fc1 = nn.Linear(input_dim, output_dim)
+        self.norm = LlamaRMSNorm(output_dim, eps=1e-6)
+        self.fc2 = nn.Linear(output_dim, output_dim)
+
+    def forward(self, features, **kwargs):    
+        x = self.fc1(features)
+        x = self.norm(x)
+        x = self.fc2(x)
+        return x
+
+
+# @auto_docstring
+class VibeVoicePreTrainedModel(PreTrainedModel):
+    config_class = VibeVoiceConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _skip_keys_device_placement = "past_key_values"
+    _supports_cache_class = True
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_quantized_cache = True
+    _supports_static_cache = True
+    _supports_attention_backend = True
+
+    def _init_weights(self, module):
+        if isinstance(module, VibeVoiceDiffusionHead):
+            module.initialize_weights()
+            return
+
+        # Use the language model's initializer_range if available
+        if hasattr(self.config, 'language_model_config') and hasattr(self.config.language_model_config, 'initializer_range'):
+            std = self.config.language_model_config.initializer_range
+        elif hasattr(self.config, 'decoder_config') and hasattr(self.config.decoder_config, 'initializer_range'):
+            std = self.config.decoder_config.initializer_range
+        else:
+            std = 0.02  # Default value
+            
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.LayerNorm):
+            module.weight.data.fill_(1.0)
+            module.bias.data.zero_()
+
+# @auto_docstring
+class VibeVoiceModel(VibeVoicePreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        
+        if hasattr(config, 'torch_dtype') and config.torch_dtype is not None:
+            if isinstance(config.torch_dtype, str):
+                dtype = getattr(torch, config.torch_dtype)
+            else:
+                dtype = config.torch_dtype
+        else:
+            dtype = torch.float32
+        
+        # Initialize Qwen2 model for language modeling
+        lm_config = config.decoder_config 
+        self.language_model = AutoModel.from_config(lm_config)
+        
+        # Initialize speech components if needed
+        self.acoustic_tokenizer = AutoModel.from_config(config.acoustic_tokenizer_config).to(dtype)
+        self.semantic_tokenizer = AutoModel.from_config(config.semantic_tokenizer_config).to(dtype)
+
+        self.acoustic_connector = SpeechConnector(config.acoustic_vae_dim, lm_config.hidden_size).to(dtype)
+        self.semantic_connector = SpeechConnector(config.semantic_vae_dim, lm_config.hidden_size).to(dtype)
+        
+        # Register scaling factors as buffers - use 1D tensors for FSDP compatibility
+        self.register_buffer('speech_scaling_factor', torch.tensor(float('nan')))  
+        self.register_buffer('speech_bias_factor', torch.tensor(float('nan')))
+
+        # Initialize prediction head for speech generation
+        self.prediction_head = AutoModel.from_config(config.diffusion_head_config).to(dtype)
+
+        # Initialize noise scheduler
+        self.noise_scheduler = DPMSolverMultistepScheduler(
+            num_train_timesteps=config.diffusion_head_config.ddpm_num_steps,
+            beta_schedule=config.diffusion_head_config.ddpm_beta_schedule,
+            prediction_type=config.diffusion_head_config.prediction_type
+        )
+    
+    def get_input_embeddings(self):
+        if hasattr(self.language_model, 'embed_tokens'):
+            # If the language model has an embed_tokens attribute, return it
+            return self.language_model.embed_tokens
+        
+        for name, attr in self.language_model.fullmap.items(): # parallel by nnscaler, the name is changed
+            if attr.orig_name == 'embed_tokens.weight':
+                return getattr(self.language_model, name)
+        assert False, 'should not arrive here'
+
+    def set_input_embeddings(self, value):
+        self.language_model.embed_tokens = value
+    
+    def set_speech_tokenizers(self, acoustic_tokenizer=None, semantic_tokenizer=None):
+        """Set the speech tokenizers used for encoding and decoding speech."""
+        self.acoustic_tokenizer = acoustic_tokenizer
+        self.semantic_tokenizer = semantic_tokenizer
+        
+        # Reset the encoder to evaluation mode
+        if self.acoustic_tokenizer is not None:
+            self.acoustic_tokenizer.eval()
+            
+        if self.semantic_tokenizer is not None:
+            self.semantic_tokenizer.eval()
+    
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs,
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        
+        # Forward through language model
+        outputs = self.language_model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            cache_position=cache_position,
+            **kwargs,
+        )
+        
+        if not return_dict:
+            return outputs
+            
+        return BaseModelOutputWithPast(
+            last_hidden_state=outputs.last_hidden_state,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+class VibeVoiceForConditionalGeneration(VibeVoicePreTrainedModel):
+    _tied_weights_keys = ["lm_head.weight"]
+    _tp_plan = {"lm_head": "colwise_rep"}
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = VibeVoiceModel(config)
+        self.vocab_size = config.decoder_config.vocab_size
+        self.lm_head = nn.Linear(config.decoder_config.hidden_size, self.vocab_size, bias=False)
+
+        self.post_init()
+        
+    def get_input_embeddings(self):
+        return self.model.get_input_embeddings()
+
+    def set_input_embeddings(self, value):
+        self.model.set_input_embeddings(value)
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_decoder(self, decoder):
+        self.model.language_model = decoder
+
+    def get_decoder(self):
+        return self.model.language_model
+
+    def tie_weights(self):
+        """
+        Tie the weights between the input embeddings and the output embeddings.
+        """
+        if getattr(self.config.decoder_config, 'tie_word_embeddings', False):
+            # The standard PreTrainedModel method will handle the tying.
+            # It typically does a simple parameter object assignment, which is
+            # CORRECT to do BEFORE FSDP wraps the model.
+            output_embeddings = self.get_output_embeddings()
+            input_embeddings = self.get_input_embeddings()
+            if hasattr(input_embeddings, 'weight'):
+                output_embeddings.weight = input_embeddings.weight
+            else:
+                # maybe returned input_embeddings a tensor directly
+                output_embeddings.weight = input_embeddings
+
+            if getattr(output_embeddings, "bias", None) is not None:
+                output_embeddings.bias.data = nn.functional.pad(
+                    output_embeddings.bias.data,
+                    (0, output_embeddings.weight.shape[0] - output_embeddings.bias.shape[0]),
+                    "constant",
+                    0,
+                )
+            print("✅ Tied input and output embeddings using standard assignment.")
+        else:
+            print("ℹ️  tie_word_embeddings is False, not tying weights.")
+
+    # Also, ensure set_output_embeddings is safe, though your implementation looks okay.
+    # The key is to avoid calling it after accelerator.prepare().
+    def set_output_embeddings(self, new_embeddings):
+        # Your current implementation using data.copy_ is good practice,
+        # but the best way is to not call this after prepare().
+        self.lm_head = new_embeddings
+
+    def forward_speech_features(
+            self, 
+            speech_tensors=None, 
+            speech_masks=None, 
+            speech_type="audio", 
+            return_unmask=False
+        ):
+        if speech_tensors is None:
+            # Use config to get vae_dim instead of non-existent self.args
+            vae_dim = self.config.acoustic_tokenizer_config.vae_dim
+            audio_features = torch.zeros(1, 1, vae_dim).to(self.get_input_embeddings().weight)
+            connect_features = self.model.acoustic_connector(audio_features)
+            return audio_features, connect_features
+        else:
+            with torch.no_grad():
+                if speech_type == "audio":
+                    with torch.no_grad():
+                        frames = self.model.acoustic_tokenizer.encode(speech_tensors.unsqueeze(1))[0][0]
+                    audio_tokens = frames.sample(self.model.acoustic_tokenizer.std_dist_type)[0]
+
+                elif speech_type == "vae":
+                    # Use config to get vae_dim instead of non-existent self.args
+                    vae_dim = self.config.acoustic_tokenizer_config.vae_dim
+                    speech_mode = speech_tensors.reshape(speech_tensors.size(0), -1, vae_dim)
+
+                    # gaussian sample from the speech_mode
+                    batch_size = speech_mode.size(0)
+                    value = self.model.acoustic_tokenizer.fix_std / 0.8
+                    std = torch.randn(batch_size, dtype=speech_mode.dtype, device=speech_mode.device) * value
+                    std = std.view(-1, *[1] * (speech_mode.dim() - 1))
+                    audio_tokens = speech_mode + std * torch.randn(speech_mode.shape).to(speech_mode)
+                else:
+                    raise NotImplementedError(f"Speech type {speech_type} not implemented")
+                
+                if torch.isnan(self.model.speech_scaling_factor) or torch.isnan(self.model.speech_bias_factor):
+                    scaling_factor = 1. / audio_tokens[speech_masks].flatten().std()
+                    bias_factor = -audio_tokens[speech_masks].flatten().mean()
+                    
+                    # Only use distributed operations if the process group is initialized
+                    if dist.is_available() and dist.is_initialized():
+                        dist.all_reduce(scaling_factor, op=dist.ReduceOp.SUM)
+                        dist.all_reduce(bias_factor, op=dist.ReduceOp.SUM)
+                        world_size = dist.get_world_size()
+                        self.model.speech_scaling_factor.copy_(scaling_factor / world_size)  
+                        self.model.speech_bias_factor.copy_(bias_factor / world_size)
+                        print(f"Speech scaling factor (distributed): {self.model.speech_scaling_factor}, bias factor: {self.model.speech_bias_factor}", flush=True)
+                    else:
+                        # Single process case
+                        self.model.speech_scaling_factor.copy_(scaling_factor)  
+                        self.model.speech_bias_factor.copy_(bias_factor)
+                        print(f"Speech scaling factor (single process): {self.model.speech_scaling_factor}, bias factor: {self.model.speech_bias_factor}", flush=True)
+                    
+                audio_features = (audio_tokens + self.model.speech_bias_factor) * self.model.speech_scaling_factor
+            
+            connect_features = self.model.acoustic_connector(audio_features)
+            if return_unmask:
+                return audio_features, connect_features
+            return audio_features[speech_masks], connect_features[speech_masks]
+        
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = False,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        # New arguments for speech processing and loss calculation
+        speech_tensors: Optional[torch.FloatTensor] = None,
+        speech_masks: Optional[torch.BoolTensor] = None,
+        speeches_loss_input: Optional[torch.FloatTensor] = None,
+        speech_semantic_tensors: Optional[torch.FloatTensor] = None, 
+        acoustic_input_mask: Optional[torch.BoolTensor] = None,
+        acoustic_loss_mask: Optional[torch.BoolTensor] = None,
+        ddpm_batch_mul: int = 1,
+        **kwargs: Optional[Dict[str, Union[torch.Tensor, str]]],
+        ) -> Union[Tuple, VibeVoiceCausalLMOutputWithPast]:
+        
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        
+        x = self.get_input_embeddings()(input_ids)
+
+        semantic_speech_all_connect_features = self.model.semantic_connector(speech_semantic_tensors)
+        if speeches_loss_input is not None:
+            # only part audio need diffuse
+            speech_all_features, speech_all_connect_features = self.forward_speech_features(
+                    speech_tensors=speech_tensors.type_as(x) if speech_tensors is not None else None,
+                    speech_masks=speech_masks,
+                    speech_type=kwargs.get("speech_type", "audio"),
+                    return_unmask=True
+                )
+            if speech_tensors is not None:
+                if semantic_speech_all_connect_features is not None:
+                    x[acoustic_input_mask] = speech_all_connect_features[speech_masks] + semantic_speech_all_connect_features[speech_masks]
+                else:
+                    x[acoustic_input_mask] = speech_all_connect_features[speech_masks]
+                speech_features = speech_all_features[speeches_loss_input.unsqueeze(-1) & speech_masks] # only part audio need diffuse
+                speech_connect_features = speech_all_connect_features[speeches_loss_input.unsqueeze(-1) & speech_masks]
+        else:
+            speech_features, speech_connect_features = self.forward_speech_features(
+                    speech_tensors=speech_tensors.type_as(x) if speech_tensors is not None else None,
+                    speech_masks=speech_masks,
+                    speech_type=kwargs.get("speech_type", "audio"),
+                )
+            if speech_tensors is not None:
+                x[acoustic_input_mask] = speech_connect_features
+
+        outputs = self.model(
+            input_ids=None,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=x,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=False,
+            return_dict=return_dict,
+            cache_position=cache_position,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        logits = self.lm_head(hidden_states)
+        # logits = logits.float()
+
+        loss = None
+        if labels is not None:
+            # The custom CE loss with masking is calculated in the training script.
+            # We leave the standard loss calculation here as None.
+            pass
+
+        # --- Diffusion Loss Calculation ---
+        diffusion_loss = None
+        # This block is executed only if we are in a context that involves speech.
+        if speech_tensors is not None and acoustic_loss_mask.sum().item() > 0:
+            condition_features = hidden_states[acoustic_loss_mask]
+            
+            speech_len, latent_size = speech_features.shape
+            
+            noise = torch.randn(
+                (speech_len * ddpm_batch_mul, latent_size),
+                device=hidden_states.device,
+                dtype=hidden_states.dtype
+            )
+            
+            timesteps = torch.multinomial(
+                torch.ones(self.config.diffusion_head_config.ddpm_num_steps),
+                speech_len * ddpm_batch_mul,
+                replacement=True,
+            ).to(hidden_states.device)
+
+            speech_features_repeated = speech_features.repeat_interleave(ddpm_batch_mul, dim=0)
+            condition_features_repeated = condition_features.repeat_interleave(ddpm_batch_mul, dim=0)
+
+            noisy_speech_features = self.model.noise_scheduler.add_noise(
+                speech_features_repeated, noise, timesteps
+            )
+            
+            model_output = self.model.prediction_head(
+                noisy_speech_features, 
+                timesteps.type_as(x), 
+                condition_features_repeated
+            )
+
+            prediction_type = self.config.diffusion_head_config.prediction_type
+            if prediction_type == "epsilon":
+                target_for_loss = noise
+            elif prediction_type == "v_prediction":
+                target_for_loss = self.model.noise_scheduler.get_velocity(
+                    speech_features_repeated, noise, timesteps
+                )
+            else:
+                raise NotImplementedError(f"Prediction type {prediction_type} not implemented")
+
+            diffusion_loss = F.mse_loss(model_output.float(), target_for_loss.float(), reduction='sum')
+            if latent_size > 0 and ddpm_batch_mul > 0:
+                diffusion_loss = diffusion_loss / latent_size / ddpm_batch_mul
+            else:
+                diffusion_loss = torch.tensor(0.0, device=diffusion_loss.device)
+        
+        else:
+            # Dummy loss for DDP to work when there are no speech samples in a batch,
+            # but we are in a speech context.
+            diffusion_loss = sum(p.sum() for p in self.model.prediction_head.parameters()) * 0.0
+            diffusion_loss += sum(p.sum() for p in self.model.acoustic_connector.parameters()) * 0.0
+            diffusion_loss += sum(p.sum() for p in self.model.semantic_connector.parameters()) * 0.0
+        # --- End Diffusion Loss Calculation ---
+
+        if not return_dict:
+            output = (logits, speech_len) + outputs.to_tuple()[1:]
+            return (loss, diffusion_loss) + output
+
+        return VibeVoiceCausalLMOutputWithPast(
+            loss=loss,
+            diffusion_loss=diffusion_loss,
+            speech_token_num=speech_len if speech_tensors is not None else 0,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+AutoModel.register(VibeVoiceConfig, VibeVoiceModel)
+AutoModelForCausalLM.register(VibeVoiceConfig, VibeVoiceForConditionalGeneration)
+
+__all__ = [
+    "VibeVoiceModel",
+    "VibeVoicePreTrainedModel",
+    "VibeVoiceForConditionalGeneration",
+    "VibeVoiceCausalLMOutputWithPast",
+    "VibeVoiceGenerationOutput",
+]

vvembed/modular/modeling_vibevoice_inference.py

@@ -0,0 +1,838 @@
+# Original code by Microsoft
+# updated by Fabio Sarracino - Enemyx-net
+
+from dataclasses import dataclass
+from typing import Dict, List, Optional, Tuple, Union, Callable
+from tqdm import tqdm
+import torch
+import torch.nn as nn
+
+from transformers.models.auto import AutoModel, AutoModelForCausalLM
+
+from transformers.generation import GenerationMixin, GenerationConfig, LogitsProcessor, LogitsProcessorList, StoppingCriteriaList
+from transformers.modeling_outputs import BaseModelOutputWithPast, ModelOutput
+from transformers import modeling_utils
+from transformers.modeling_utils import PreTrainedModel
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.utils import logging
+
+
+# from .modular_vibevoice_tokenizer import VibeVoiceTokenizerStreamingCache, VibeVoiceAcousticTokenizerModel, VibeVoiceSemanticTokenizerModel
+from .modular_vibevoice_tokenizer import VibeVoiceTokenizerStreamingCache, VibeVoiceTokenizerEncoderOutput
+from .modular_vibevoice_diffusion_head import VibeVoiceDiffusionHead
+from schedule.dpm_solver import DPMSolverMultistepScheduler
+
+from .configuration_vibevoice import VibeVoiceConfig
+
+from .modular_vibevoice_text_tokenizer import VibeVoiceTextTokenizer, VibeVoiceTextTokenizerFast
+
+from .modeling_vibevoice import VibeVoiceModel, VibeVoicePreTrainedModel
+from .streamer import AudioStreamer, AsyncAudioStreamer
+
+logger = logging.get_logger(__name__)
+
+if not hasattr(modeling_utils, "ALL_PARALLEL_STYLES") or modeling_utils.ALL_PARALLEL_STYLES is None:
+    modeling_utils.ALL_PARALLEL_STYLES = ["tp", "none", "colwise", "rowwise"]
+
+@dataclass
+class VibeVoiceCausalLMOutputWithPast(BaseModelOutputWithPast):
+    logits: Optional[torch.FloatTensor] = None
+
+@dataclass
+class VibeVoiceGenerationOutput(ModelOutput):
+    """
+    Output type for VibeVoice generation.
+    
+    Args:
+        sequences (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            The generated sequences. 
+        speech_outputs (`List[torch.FloatTensor]`, *optional*):
+            List of generated speech waveforms or latents for each speech segment.
+    """
+    sequences: torch.LongTensor = None
+    speech_outputs: Optional[List[torch.FloatTensor]] = None
+    reach_max_step_sample: Optional[torch.BoolTensor] = None
+
+class VibeVoiceTokenConstraintProcessor(LogitsProcessor):
+    """Constrains token generation to only valid tokens during speech generation."""
+    
+    def __init__(self, valid_token_ids: List[int], device: torch.device = None):
+        self.valid_token_ids = torch.tensor(valid_token_ids, dtype=torch.long, device=device)
+        
+    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
+        # Create a mask for valid tokens
+        mask = torch.full_like(scores, float('-inf'))
+        mask[:, self.valid_token_ids] = 0
+        
+        # Apply mask to scores
+        scores = scores + mask
+        return scores
+
+def access_cache_safely(cache, layer_idx):
+    """Access cache tensors safely across different transformers versions
+    
+    This function handles the different DynamicCache structures across transformers versions:
+    - Old versions (< 4.36): cache.key_cache, cache.value_cache
+    - Intermediate versions: cache._keys, cache._values  
+    - New versions (4.36+): Various new structures
+    
+    Returns (k_cache, v_cache) or (None, None) if cache structure is incompatible
+    """
+    try:
+        # Method 1: Old versions (< 4.36)
+        if hasattr(cache, 'key_cache') and hasattr(cache, 'value_cache'):
+            if layer_idx < len(cache.key_cache):
+                return cache.key_cache[layer_idx], cache.value_cache[layer_idx]
+        
+        # Method 2: Private attributes (some intermediate versions)
+        if hasattr(cache, '_keys') and hasattr(cache, '_values'):
+            if layer_idx < len(cache._keys):
+                return cache._keys[layer_idx], cache._values[layer_idx]
+        
+        # Method 3: New versions with get_seq_length or similar
+        # Some versions store as list of tuples
+        if isinstance(cache, (list, tuple)) and len(cache) > layer_idx:
+            layer_cache = cache[layer_idx]
+            if isinstance(layer_cache, (list, tuple)) and len(layer_cache) >= 2:
+                return layer_cache[0], layer_cache[1]
+            elif hasattr(layer_cache, 'key_states') and hasattr(layer_cache, 'value_states'):
+                return layer_cache.key_states, layer_cache.value_states
+        
+        # Method 4: Check if cache has a different structure entirely
+        # Some very new versions might not expose cache directly
+        if hasattr(cache, 'to_legacy_tuple'):
+            # Convert to legacy format if possible
+            legacy = cache.to_legacy_tuple()
+            if legacy and layer_idx < len(legacy):
+                return legacy[layer_idx][0], legacy[layer_idx][1]
+                
+    except (AttributeError, IndexError, TypeError) as e:
+        # Log the issue but don't fail
+        logger.debug(f"Could not access cache at layer {layer_idx}: {e}")
+    
+    # Return None if we can't access the cache safely
+    return None, None
+
+def get_num_layers_from_cache(cache):
+    """Get the number of layers in the cache structure"""
+    try:
+        if hasattr(cache, 'key_cache'):
+            return len(cache.key_cache)
+        elif hasattr(cache, '_keys'):
+            return len(cache._keys)
+        elif isinstance(cache, (list, tuple)):
+            return len(cache)
+        elif hasattr(cache, 'num_layers'):
+            return cache.num_layers
+        # Default fallback - most models have 32 or fewer layers
+        return 32
+    except:
+        return 32
+    
+class VibeVoiceForConditionalGenerationInference(VibeVoicePreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["lm_head.weight"]
+    _tp_plan = {"lm_head": "colwise_rep"}
+
+    def __init__(self, config):
+        super().__init__(config)
+        
+        # Initialize the base model
+        self.model = VibeVoiceModel(config)
+        
+        # LM head for text generation
+        self.lm_head = nn.Linear(config.decoder_config.hidden_size, config.decoder_config.vocab_size, bias=False)
+        
+        # inference configuration
+        self.ddpm_inference_steps = config.diffusion_head_config.ddpm_num_inference_steps
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @property
+    def noise_scheduler(self):
+        return self.model.noise_scheduler
+
+    @property
+    def prediction_head(self):
+        return self.model.prediction_head
+    
+    @property
+    def speech_scaling_factor(self):
+        return self.model.speech_scaling_factor
+
+    @property
+    def speech_bias_factor(self):
+        return self.model.speech_bias_factor
+
+    @property
+    def acoustic_tokenizer(self):
+        return self.model.acoustic_tokenizer
+
+    @property
+    def semantic_tokenizer(self):
+        return self.model.semantic_tokenizer
+    
+    @property
+    def acoustic_connector(self):
+        return self.model.acoustic_connector
+
+    @property
+    def semantic_connector(self):
+        return self.model.semantic_connector
+        
+    def tie_weights(self):
+        """
+        Tie the weights between the input embeddings and the output embeddings.
+        """
+        # Tie lm_head.weight to language_model.embed_tokens.weight
+        if not getattr(self.config, 'tie_word_embeddings', False):
+            return
+         
+        if hasattr(self, 'lm_head') and hasattr(self.model.language_model, 'embed_tokens'):
+            self.lm_head.weight = self.model.language_model.embed_tokens.weight
+        
+    def get_input_embeddings(self):
+        return self.model.get_input_embeddings()
+    
+    def set_input_embeddings(self, value):
+        self.model.set_input_embeddings(value)
+    
+    def get_output_embeddings(self):
+        return self.lm_head
+    
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+    
+    def set_speech_tokenizers(self, acoustic_tokenizer=None, semantic_tokenizer=None):
+        """Set the speech tokenizers used for encoding and decoding speech."""
+        self.model.set_speech_tokenizers(acoustic_tokenizer, semantic_tokenizer)
+    
+    def set_ddpm_inference_steps(self, num_steps=None):
+        self.ddpm_inference_steps = num_steps or self.config.diffusion_head_config.ddpm_num_inference_steps
+
+    def _process_speech_inputs(self, speech_tensors, speech_masks, speech_type="audio"):
+        """Process speech inputs through tokenizers and connectors."""
+        with torch.no_grad():
+            if speech_type == "audio":
+                # Encode audio to acoustic latents
+                encoder_output = self.model.acoustic_tokenizer.encode(speech_tensors.unsqueeze(1))
+                acoustic_latents = encoder_output.sample(dist_type=self.model.acoustic_tokenizer.std_dist_type)[0]
+                
+                # Apply scaling and bias
+                acoustic_features = (acoustic_latents + self.model.speech_bias_factor.to(acoustic_latents.device)) * self.model.speech_scaling_factor.to(acoustic_latents.device)
+                
+                # Connect to language model space
+                acoustic_connected = self.model.acoustic_connector(acoustic_features)[speech_masks.cpu()]
+                
+                return acoustic_features, acoustic_connected
+            elif speech_type == "pt":
+                encoder_output = VibeVoiceTokenizerEncoderOutput(mean=speech_tensors, std=self.acoustic_tokenizer.config.fix_std)
+                acoustic_latents = encoder_output.sample(dist_type=self.model.acoustic_tokenizer.std_dist_type)[0]
+                
+                # Apply scaling and bias
+                acoustic_features = (acoustic_latents + self.model.speech_bias_factor.to(acoustic_latents.device)) * self.model.speech_scaling_factor.to(acoustic_latents.device)
+                
+                # Connect to language model space
+                acoustic_connected = self.model.acoustic_connector(acoustic_features)[speech_masks.cpu()]
+                
+                return acoustic_features, acoustic_connected
+            else:
+                raise NotImplementedError(f"Speech type {speech_type} not implemented")
+    
+    # @can_return_tuple
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        speech_tensors: Optional[torch.FloatTensor] = None,
+        speech_masks: Optional[torch.BoolTensor] = None,
+        speech_input_mask: Optional[torch.BoolTensor] = None,
+        logits_to_keep: Union[int, slice] = 0,
+        **kwargs,
+    ) -> Union[Tuple, VibeVoiceCausalLMOutputWithPast]:
+        """
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+            speech_tensors (`torch.FloatTensor`, *optional*):
+                Input speech waveforms for voice cloning or speech understanding.
+            speech_masks (`torch.BoolTensor`, *optional*):
+                Masks indicating valid speech frames.
+            speech_input_mask (`torch.BoolTensor`, *optional*):
+                Positions in the input sequence where speech embeddings should be inserted.
+        
+        Returns:
+            `VibeVoiceCausalLMOutputWithPast` or tuple
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        
+        # Get embeddings
+        if inputs_embeds is None:
+            inputs_embeds = self.model.get_input_embeddings()(input_ids)
+        
+        # Process speech inputs if provided
+        if speech_tensors is not None and speech_masks is not None:
+            acoustic_features, speech_embeds = self._process_speech_inputs(speech_tensors.to(self.dtype), speech_masks)
+            if speech_input_mask is not None:
+                inputs_embeds[speech_input_mask] = speech_embeds
+
+        outputs = self.model(
+            inputs_embeds=inputs_embeds,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs[0] if not return_dict else outputs.last_hidden_state
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+                
+        if labels is not None:
+            raise NotImplementedError("Loss computation is not implemented in this version.")
+
+        return VibeVoiceCausalLMOutputWithPast(
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            last_hidden_state=hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def _build_generate_config_model_kwargs(self, generation_config, inputs, tokenizer, return_processors=False, **kwargs):
+        if generation_config is None:
+            generation_config = GenerationConfig(
+                bos_token_id=tokenizer.bos_token_id,
+                eos_token_id=tokenizer.eos_token_id,
+                pad_token_id = tokenizer.pad_token_id
+            )
+        else:
+            generation_config = GenerationConfig(
+                **generation_config,
+                bos_token_id=tokenizer.bos_token_id,
+                eos_token_id=tokenizer.eos_token_id,
+                pad_token_id = tokenizer.pad_token_id
+            )
+
+        generation_config, model_kwargs = self._prepare_generation_config(
+            generation_config, 
+            True, 
+            speech_start_id=tokenizer.speech_start_id, 
+            speech_end_id=tokenizer.speech_end_id, 
+            speech_diffusion_id=tokenizer.speech_diffusion_id, 
+            **kwargs
+        )
+        generation_config.speech_start_id = tokenizer.speech_start_id
+        generation_config.speech_end_id = tokenizer.speech_end_id
+        generation_config.speech_diffusion_id = tokenizer.speech_diffusion_id
+
+        inputs_tensor, model_input_name, model_kwargs = self._prepare_model_inputs(inputs, generation_config.bos_token_id, model_kwargs)
+        batch_size = inputs_tensor.shape[0]
+        device = self.device
+        
+        self._prepare_special_tokens(generation_config, True, device=device)
+        generation_config.use_cache = True
+        model_kwargs["use_cache"] = generation_config.use_cache
+        input_ids = inputs_tensor.to(self.device)
+
+        input_ids_length = input_ids.shape[1]
+        has_default_max_length = kwargs.get("max_length") is None and generation_config.max_length is not None
+        has_default_min_length = kwargs.get("min_length") is None and generation_config.min_length is not None
+        generation_config = self._prepare_generated_length(
+            generation_config=generation_config,
+            has_default_max_length=has_default_max_length,
+            has_default_min_length=has_default_min_length,
+            model_input_name=model_input_name,
+            inputs_tensor=inputs_tensor,
+            input_ids_length=input_ids_length,
+        )
+
+        max_cache_length = generation_config.max_length - 1
+        
+        # Fix for transformers compatibility: detect number of parameters accepted
+        import inspect
+        try:
+            sig = inspect.signature(self._prepare_cache_for_generation)
+            num_params = len(sig.parameters)
+            
+            # Newer transformers expects 6 parameters (without 'device')
+            # Older transformers expects 7 parameters (with 'device')
+            if num_params == 6 or 'device' not in sig.parameters:
+                # New signature: (self, generation_config, model_kwargs, assistant_model, batch_size, max_cache_length)
+                self._prepare_cache_for_generation(generation_config, model_kwargs, None, batch_size, max_cache_length)
+            else:
+                # Old signature: (self, generation_config, model_kwargs, assistant_model, batch_size, max_cache_length, device)
+                self._prepare_cache_for_generation(generation_config, model_kwargs, None, batch_size, max_cache_length, device)
+        except Exception as e:
+            # Fallback: try both signatures
+            try:
+                # Try new signature first (6 params)
+                self._prepare_cache_for_generation(generation_config, model_kwargs, None, batch_size, max_cache_length)
+            except TypeError:
+                # Fall back to old signature (7 params)
+                self._prepare_cache_for_generation(generation_config, model_kwargs, None, batch_size, max_cache_length, device)
+        
+        model_kwargs['cache_position'] = torch.arange(input_ids_length, device=device, dtype=torch.long)
+        for k, v in model_kwargs.items():
+            if isinstance(v, torch.Tensor):
+                model_kwargs[k] = v.to(device=device)
+        
+        if return_processors:
+            logits_processor = self._get_logits_processor(
+                generation_config=generation_config,
+                input_ids_seq_length=input_ids_length,
+                encoder_input_ids=inputs_tensor,
+                prefix_allowed_tokens_fn=None,
+                logits_processor=LogitsProcessorList(),
+                device=inputs_tensor.device,
+                model_kwargs=model_kwargs,
+            )
+
+            stopping_criteria = self._get_stopping_criteria(generation_config=generation_config, stopping_criteria=StoppingCriteriaList())
+        
+            return generation_config, model_kwargs, input_ids, logits_processor, stopping_criteria
+        else:
+            return generation_config, model_kwargs, input_ids
+        
+    @torch.no_grad()
+    def generate(
+        self,
+        inputs: Optional[torch.Tensor] = None,
+        generation_config: Optional[GenerationConfig] = None,
+        logits_processor: Optional[LogitsProcessorList] = None,
+        stopping_criteria: Optional[StoppingCriteriaList] = None,
+        prefix_allowed_tokens_fn: Optional[Callable[[int, torch.Tensor], List[int]]] = None,
+        synced_gpus: Optional[bool] = None,
+        assistant_model: Optional["PreTrainedModel"] = None,
+        audio_streamer: Optional[Union[AudioStreamer, AsyncAudioStreamer]] = None,
+        negative_prompt_ids: Optional[torch.Tensor] = None,
+        negative_prompt_attention_mask: Optional[torch.Tensor] = None,
+        speech_tensors: Optional[torch.FloatTensor] = None,
+        speech_masks: Optional[torch.BoolTensor] = None,
+        speech_input_mask: Optional[torch.BoolTensor] = None,
+        return_speech: bool = True,
+        cfg_scale: float = 1.0,
+        stop_check_fn: Optional[Callable[[], bool]] = None,
+        **kwargs,
+    ) -> Union[torch.LongTensor, VibeVoiceGenerationOutput]:
+        """
+        Generates sequences of token ids and optionally speech outputs.
+        
+        Args:
+            All standard generation arguments from GenerationMixin
+            negative_prompt_ids: Negative prompt for CFG in speech generation
+            negative_prompt_attention_mask: Attention mask for negative prompt
+            speech_tensors: Input speech for voice cloning
+            speech_masks: Masks for speech tensors  
+            speech_input_mask: Positions to insert speech embeddings
+            return_speech: Whether to decode and return speech outputs
+            cfg_scale: CFG scale for speech generation
+            stop_check_fn: Optional callable that returns True if generation should stop
+ 
+        Returns:
+            Generated token sequences and optionally speech outputs
+        """
+        # 1. Handle `generation_config` and kwargs that might update it, and validate the `.generate()` call
+        tokenizer = kwargs.pop("tokenizer", None)  # Pull this out first, we only use it for stopping criteria
+        parsed_scripts = kwargs.pop("parsed_scripts", None)
+        all_speakers_list = kwargs.pop("all_speakers_list", None)
+        max_length_times = kwargs.pop("max_length_times", 2)
+
+        if kwargs.get('max_new_tokens', None) is None:
+            kwargs['max_new_tokens'] = self.config.decoder_config.max_position_embeddings - kwargs['input_ids'].shape[-1]
+
+        generation_config, model_kwargs, input_ids, logits_processor, stopping_criteria = self._build_generate_config_model_kwargs(
+            generation_config, inputs, tokenizer, return_processors=True, **kwargs
+        )
+        
+        negative_kwargs = {
+            'input_ids': torch.full((kwargs['input_ids'].shape[0], 1), tokenizer.speech_start_id, dtype=torch.long, device=kwargs['input_ids'].device),
+            'attention_mask':  torch.ones((kwargs['input_ids'].shape[0], 1), dtype=torch.long, device=kwargs['input_ids'].device),
+            'max_new_tokens': kwargs.get('max_new_tokens', 100) 
+        }
+        negative_generation_config, negative_model_kwargs, negative_input_ids = self._build_generate_config_model_kwargs(
+            None, None, tokenizer, return_processors=False, **negative_kwargs
+        )
+
+        acoustic_cache = VibeVoiceTokenizerStreamingCache()
+        semantic_cache = VibeVoiceTokenizerStreamingCache()
+        
+        batch_size = input_ids.shape[0]
+        device = input_ids.device
+        finished_tags = torch.zeros(batch_size, dtype=torch.bool, device=device)
+        correct_cnt = torch.zeros(batch_size, dtype=torch.long, device=device)
+        is_prefill = True
+        inputs_embeds = None
+        verbose = kwargs.get("verbose", False)
+
+        # Initialize audio chunks storage for each sample
+        audio_chunks = [[] for _ in range(batch_size)]
+
+        initial_length = input_ids.shape[-1]
+        initial_length_per_sample = model_kwargs['attention_mask'].sum(dim=-1)
+
+       # Define all valid tokens that can be generated
+        valid_tokens = [
+            generation_config.speech_start_id,
+            generation_config.speech_end_id, 
+            generation_config.speech_diffusion_id,
+            generation_config.eos_token_id
+        ]
+        # Add bos_token_id if it exists
+        if hasattr(generation_config, 'bos_token_id') and generation_config.bos_token_id is not None:
+            valid_tokens.append(generation_config.bos_token_id)
+        
+        # Add custom processor to constrain token generation
+        token_constraint_processor = VibeVoiceTokenConstraintProcessor(valid_tokens, device=device)
+        if logits_processor is None:
+            logits_processor = LogitsProcessorList()
+        logits_processor.append(token_constraint_processor)
+        
+        max_steps = min(generation_config.max_length - initial_length, int(max_length_times * initial_length))
+        max_step_per_sample = torch.min(generation_config.max_length - initial_length_per_sample, (max_length_times * initial_length_per_sample).long())
+        reach_max_step_sample = torch.zeros(batch_size, dtype=torch.bool, device=device)
+
+        # Create progress iterator if verbose
+        if kwargs.get("show_progress_bar", True):
+            progress_bar = tqdm(range(max_steps), desc="Generating", leave=False)
+        else:
+            progress_bar = range(max_steps)
+        
+        for step in progress_bar:
+            # Check for external stop signal
+            if stop_check_fn is not None and stop_check_fn():
+                if verbose:
+                    print(f"Generation stopped externally at step {step + 1}")
+                # End the audio streamer if it exists
+                if audio_streamer is not None:
+                    audio_streamer.end()
+                break
+            
+            # Check if audio_streamer has been ended (stopped externally)
+            if audio_streamer is not None and hasattr(audio_streamer, 'finished_flags'):
+                if any(audio_streamer.finished_flags):
+                    if verbose:
+                        print(f"Audio generation stopped externally at step {step + 1}")
+                    break
+            
+            if finished_tags.all():
+                if hasattr(progress_bar, 'set_description'):
+                    progress_bar.set_description("Generation complete")
+                break
+
+            if input_ids.shape[-1] >= generation_config.max_length:
+                print(f"Reached maximum generation length {generation_config.max_length}, stopped it.")
+                reached_samples = torch.arange(batch_size, device=device)[~finished_tags]
+                if reached_samples.numel() > 0:
+                    reach_max_step_sample[reached_samples] = True
+                break
+            
+            # Update progress bar description with active samples
+            if hasattr(progress_bar, 'set_description'):
+                active_samples = (~finished_tags).sum().item()
+                progress_bar.set_description(f"Generating (active: {active_samples}/{batch_size})")
+
+            model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)
+            if is_prefill:
+                # we process the speech inputs only during the first generation step
+                prefill_inputs = {
+                    "speech_tensors": speech_tensors.to(device=device),
+                    "speech_masks": speech_masks.to(device),
+                    "speech_input_mask": speech_input_mask.to(device),
+                }
+                is_prefill = False
+            else:
+                _ = model_inputs.pop('inputs_embeds', None)
+                prefill_inputs = {'inputs_embeds': inputs_embeds}
+
+            # Forward pass through the model
+            outputs = self(
+                **model_inputs, **prefill_inputs, logits_to_keep=1, return_dict=True, output_attentions=False, output_hidden_states=False,
+            )
+            model_kwargs = self._update_model_kwargs_for_generation(
+                outputs, model_kwargs, is_encoder_decoder=False,
+            )
+
+            # Get logits and apply logits processor
+            next_token_logits = outputs.logits[:, -1, :].to(copy=True, dtype=torch.float32, device=input_ids.device)
+            # next_token_logits = outputs.logits[:, -1, :].to(copy=True, device=input_ids.device)
+            next_token_scores = logits_processor(input_ids, next_token_logits)
+            
+            # token selection
+            if generation_config.do_sample:
+                probs = nn.functional.softmax(next_token_scores, dim=-1)
+                # TODO (joao): this OP throws "skipping cudagraphs due to ['incompatible ops']", find solution
+                next_tokens = torch.multinomial(probs, num_samples=1).squeeze(1)
+            else:
+                next_tokens = torch.argmax(next_token_scores, dim=-1)
+
+            next_tokens[finished_tags] = generation_config.eos_token_id
+            input_ids = torch.cat([input_ids, next_tokens[:, None]], dim=-1)
+            
+            if not kwargs.get('refresh_negative', True):
+                negative_model_inputs = self.prepare_inputs_for_generation(negative_input_ids, **negative_model_kwargs)
+                # Forward negative pass through the model
+                if negative_model_inputs['inputs_embeds'] is None and inputs_embeds is not None:
+                    negative_model_inputs['inputs_embeds'] = inputs_embeds
+                    negative_model_inputs['input_ids'] = None
+
+                negative_outputs = self(
+                    **negative_model_inputs, logits_to_keep=0, return_dict=True, output_attentions=False, output_hidden_states=False,
+                )
+                negative_model_kwargs = self._update_model_kwargs_for_generation(
+                    negative_outputs, negative_model_kwargs, is_encoder_decoder=False,
+                )
+                negative_input_ids = torch.cat([negative_input_ids, next_tokens[:, None]], dim=-1)
+
+            # reached end of generation
+            if (next_tokens == generation_config.eos_token_id).any():
+                eos_indices = (next_tokens == generation_config.eos_token_id).nonzero(as_tuple=False).squeeze(1)
+                # Only print for samples that are newly finished (not already marked as finished)
+                new_eos_indices = eos_indices[~finished_tags[eos_indices]]
+                if new_eos_indices.numel() > 0:
+                    finished_tags[new_eos_indices] = True
+                    if verbose:
+                        print(f"Samples {new_eos_indices.tolist()} reached EOS token at step {step + 1}.", flush=True)
+                    if audio_streamer is not None:
+                        audio_streamer.end(new_eos_indices)
+
+            # Check if any sample reached its maximum generation length
+            max_length_reached = step >= max_step_per_sample
+            new_max_length_indices = torch.nonzero(max_length_reached & ~finished_tags, as_tuple=False).squeeze(1)
+            if new_max_length_indices.numel() > 0:
+                finished_tags[new_max_length_indices] = True
+                reach_max_step_sample[new_max_length_indices] = True
+                if verbose:
+                    print(f"Samples {new_max_length_indices.tolist()} reached max generation length at step {step + 1}.", flush=True)
+                if audio_streamer is not None:
+                    audio_streamer.end(new_max_length_indices)
+
+            # speech_end
+            diffusion_end_indices = (next_tokens == generation_config.speech_end_id).nonzero(as_tuple=False).squeeze(1)
+            if diffusion_end_indices.numel() > 0:
+                # Clear tokenizer caches for samples that reached speech end
+                acoustic_cache.set_to_zero(diffusion_end_indices)
+                semantic_cache.set_to_zero(diffusion_end_indices)
+            
+            # speech_begin
+            diffusion_start_indices = torch.arange(batch_size, device=device)[~finished_tags & (next_tokens == generation_config.speech_start_id)]
+            if diffusion_start_indices.numel() > 0 and kwargs.get('refresh_negative', True):
+                # update attention mask
+                for i, sample_idx in enumerate(diffusion_start_indices.tolist()):
+                    negative_model_kwargs['attention_mask'][sample_idx, :] = 0
+                    negative_model_kwargs['attention_mask'][sample_idx, -1] = 1
+                # update past key values - using safe cache access
+                cache = negative_model_kwargs['past_key_values']
+                num_layers = get_num_layers_from_cache(cache)
+                cache_update_failed = False
+                
+                for layer_idx in range(num_layers):
+                    k_cache, v_cache = access_cache_safely(cache, layer_idx)
+                    if k_cache is None or v_cache is None:
+                        # Cache structure not compatible, skip optimization
+                        logger.debug(f"Cache optimization skipped at layer {layer_idx} - incompatible structure")
+                        cache_update_failed = True
+                        break
+                    
+                    # Process each non-diffusion sample
+                    for sample_idx in diffusion_start_indices.tolist():
+                        try:
+                            # Shift cache for this sample
+                            k_cache[sample_idx, :, -1, :] = k_cache[sample_idx, :, 0, :].clone()
+                            v_cache[sample_idx, :, -1, :] = v_cache[sample_idx, :, 0, :].clone()
+                        except (IndexError, RuntimeError) as e:
+                            logger.debug(f"Cache update failed for sample {sample_idx}: {e}")
+                            cache_update_failed = True
+                            break
+                    
+                    if cache_update_failed:
+                        break
+                # update negative_input_ids
+                for sample_idx in diffusion_start_indices.tolist():
+                    negative_input_ids[sample_idx, -1] = generation_config.speech_start_id
+            
+            # Prepare inputs_embeds for next iteration
+            # Initialize with default embeddings for all tokens
+            next_inputs_embeds = self.model.get_input_embeddings()(next_tokens).unsqueeze(1)  # [batch_size, 1, hidden_size]
+            
+            # forward diffusion
+            # Diffusion indices are those that are not finished and not special tokens
+            diffusion_indices = torch.arange(batch_size, device=device)[~finished_tags & (next_tokens == generation_config.speech_diffusion_id)]
+            
+            if diffusion_indices.numel() > 0:
+                if kwargs.get('refresh_negative', True):
+                    negative_model_inputs = self.prepare_inputs_for_generation(negative_input_ids, **negative_model_kwargs)
+                    # Forward negative pass through the model
+                    if negative_model_inputs['inputs_embeds'] is None and inputs_embeds is not None:
+                        negative_model_inputs['inputs_embeds'] = inputs_embeds
+                        negative_model_inputs['input_ids'] = None
+
+                    negative_outputs = self(
+                        **negative_model_inputs, logits_to_keep=0, return_dict=True, output_attentions=False, output_hidden_states=False,
+                    )
+                    negative_model_kwargs = self._update_model_kwargs_for_generation(
+                        negative_outputs, negative_model_kwargs, is_encoder_decoder=False,
+                    )
+                    negative_input_ids = torch.cat([negative_input_ids, next_tokens[:, None]], dim=-1)
+                # correct the non-diffusion indices
+                # we forward all samples' negative outputs even if 
+                #   they are not in diffusion mode to keep the cache consistent
+                # So we need to correct the kv cache of non-diffusion samples
+                non_diffusion_mask = ~finished_tags & (next_tokens != generation_config.speech_diffusion_id)
+                if non_diffusion_mask.any():
+                    non_diffusion_indices = torch.arange(batch_size, device=device)[non_diffusion_mask]
+                    start_indices = correct_cnt[non_diffusion_indices]
+
+                    # 1. Update attention_mask - need to handle each sample separately
+                    seq_len = negative_model_kwargs['attention_mask'].shape[1]
+                    for i, (sample_idx, start_idx) in enumerate(zip(non_diffusion_indices.tolist(), start_indices.tolist())):
+                        # Shift the attention mask for this sample
+                        if start_idx + 1 < seq_len - 1:
+                            negative_model_kwargs['attention_mask'][sample_idx, start_idx+1:] = \
+                                negative_model_kwargs['attention_mask'][sample_idx, start_idx:-1].clone()
+                        negative_model_kwargs['attention_mask'][sample_idx, start_idx] = 0
+
+                    # 2. Update past_key_values - using safe cache access
+                    cache = negative_model_kwargs['past_key_values']
+                    num_layers = get_num_layers_from_cache(cache)
+                    cache_update_failed = False
+                    
+                    for layer_idx in range(num_layers):
+                        k_cache, v_cache = access_cache_safely(cache, layer_idx)
+                        if k_cache is None or v_cache is None:
+                            # Cache structure not compatible, skip optimization
+                            logger.debug(f"Cache optimization skipped at layer {layer_idx} - incompatible structure")
+                            cache_update_failed = True
+                            break
+                        
+                        # Process each non-diffusion sample
+                        for sample_idx, start_idx in zip(non_diffusion_indices.tolist(), start_indices.tolist()):
+                            try:
+                                if start_idx + 1 < k_cache.shape[2] - 1:
+                                    # Shift cache for this sample
+                                    k_cache[sample_idx, :, start_idx+1:, :] = k_cache[sample_idx, :, start_idx:-1, :].clone()
+                                    v_cache[sample_idx, :, start_idx+1:, :] = v_cache[sample_idx, :, start_idx:-1, :].clone()
+                            except (IndexError, RuntimeError) as e:
+                                logger.debug(f"Cache update failed for sample {sample_idx}: {e}")
+                                cache_update_failed = True
+                                break
+                        
+                        if cache_update_failed:
+                            break
+                    
+                    # 3. Update negative_input_ids
+                    for sample_idx, start_idx in zip(non_diffusion_indices.tolist(), start_indices.tolist()):
+                        if start_idx + 1 < negative_input_ids.shape[1] - 1:
+                            negative_input_ids[sample_idx, start_idx+1:] = \
+                                negative_input_ids[sample_idx, start_idx:-1].clone()
+                                
+                    correct_cnt[non_diffusion_indices] += 1
+
+                positive_condition = outputs.last_hidden_state[diffusion_indices, -1, :]
+                negative_condition = negative_outputs.last_hidden_state[diffusion_indices, -1, :]
+                
+                speech_latent = self.sample_speech_tokens(
+                    positive_condition,
+                    negative_condition,
+                    cfg_scale=cfg_scale,
+                ).unsqueeze(1)
+                                
+                # Decode acoustic latent to audio using acoustic streaming cache
+                scaled_latent = speech_latent / self.model.speech_scaling_factor.to(speech_latent.device) - self.model.speech_bias_factor.to(speech_latent.device)
+                audio_chunk = self.model.acoustic_tokenizer.decode(
+                    scaled_latent.to(self.model.acoustic_tokenizer.device),
+                    cache=acoustic_cache,  # Use acoustic-specific cache
+                    sample_indices=diffusion_indices.to(self.model.acoustic_tokenizer.device),
+                    use_cache=True,
+                    debug=False
+                )
+                
+                # Store audio chunks for each sample
+                for i, sample_idx in enumerate(diffusion_indices):
+                    idx = sample_idx.item()
+                    # Only append audio chunk if the sample is not finished
+                    if not finished_tags[idx]:
+                        audio_chunks[idx].append(audio_chunk[i])
+
+                 # Add streaming support here
+                if audio_streamer is not None:
+                    # Stream the audio chunks immediately
+                    audio_streamer.put(audio_chunk, diffusion_indices)
+                    
+                # Encode audio to semantic features using semantic streaming cache
+                semantic_features = self.model.semantic_tokenizer.encode(
+                    audio_chunk,
+                    cache=semantic_cache,  # Use semantic-specific cache
+                    sample_indices=diffusion_indices,
+                    use_cache=True,
+                    debug=False
+                ).mean # semantic tokenizer has no VAE.
+                
+                # Combine acoustic and semantic features for next input
+                acoustic_embed = self.model.acoustic_connector(speech_latent)
+                semantic_embed = self.model.semantic_connector(semantic_features)
+                diffusion_embeds = acoustic_embed + semantic_embed
+
+                # Update embeddings for diffusion indices
+                next_inputs_embeds[diffusion_indices] = diffusion_embeds
+            
+            # Set inputs_embeds for next iteration
+            inputs_embeds = next_inputs_embeds
+
+        if audio_streamer is not None:
+            audio_streamer.end()
+
+        # Concatenate audio chunks for each sample
+        final_audio_outputs = []
+        for sample_chunks in audio_chunks:
+            if sample_chunks:
+                # Concatenate all chunks along the time dimension (assumed to be the last dimension)
+                concatenated_audio = torch.cat(sample_chunks, dim=-1)
+                final_audio_outputs.append(concatenated_audio)
+            else:
+                # If no audio was generated for this sample, append None
+                final_audio_outputs.append(None)
+
+        return VibeVoiceGenerationOutput(
+            sequences=input_ids,
+            speech_outputs=final_audio_outputs if return_speech else None,
+            reach_max_step_sample=reach_max_step_sample,
+        )
+    
+    @torch.no_grad()
+    def sample_speech_tokens(self, condition, neg_condition, cfg_scale=3.0):
+        self.model.noise_scheduler.set_timesteps(self.ddpm_inference_steps)
+        condition = torch.cat([condition, neg_condition], dim=0).to(self.model.prediction_head.device)
+        speech = torch.randn(condition.shape[0], self.config.acoustic_vae_dim).to(condition)
+        for t in self.model.noise_scheduler.timesteps:
+            half = speech[: len(speech) // 2]
+            combined = torch.cat([half, half], dim=0)
+            eps = self.model.prediction_head(combined, t.repeat(combined.shape[0]).to(combined), condition=condition)
+            cond_eps, uncond_eps = torch.split(eps, len(eps) // 2, dim=0)
+            half_eps = uncond_eps + cfg_scale * (cond_eps - uncond_eps)
+            eps = torch.cat([half_eps, half_eps], dim=0)
+            speech = self.model.noise_scheduler.step(eps, t, speech).prev_sample
+        return speech[: len(speech) // 2]
+    
+
+AutoModelForCausalLM.register(VibeVoiceConfig, VibeVoiceForConditionalGenerationInference)
+
+__all__ = [
+    "VibeVoiceForConditionalGenerationInference",
+]

vvembed/modular/modular_vibevoice_diffusion_head.py

@@ -0,0 +1,287 @@
+import math
+from typing import Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from transformers.models.auto import AutoModel
+from transformers.modeling_utils import PreTrainedModel
+# from transformers.modeling_layers import GradientCheckpointingLayer
+from transformers.activations import ACT2FN
+from transformers.utils import logging
+
+from .configuration_vibevoice import VibeVoiceDiffusionHeadConfig
+
+
+logger = logging.get_logger(__name__)
+
+
+class RMSNorm(nn.Module):
+    def __init__(self, dim: int, eps: float = 1e-6, elementwise_affine=True, memory_efficient=False):
+        super().__init__()
+        self.dim = dim
+        self.eps = eps
+        self.elementwise_affine = elementwise_affine
+        if self.elementwise_affine:
+            self.weight = nn.Parameter(torch.ones(dim))
+        else:
+            self.register_parameter('weight', None)
+
+    def _norm(self, x):
+        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)
+
+    def forward(self, x):
+        output = self._norm(x.float()).type_as(x)
+        if self.weight is not None:
+            output = output * self.weight
+        return output
+
+    def extra_repr(self) -> str:
+        return f'dim={self.dim}, eps={self.eps}, elementwise_affine={self.elementwise_affine}'
+    
+def modulate(x, shift, scale):
+    """Apply modulation to input tensor."""
+    return x * (1 + scale) + shift
+
+
+class TimestepEmbedder(nn.Module):
+    """
+    Embeds scalar timesteps into vector representations.
+    
+    Args:
+        hidden_size (`int`): Size of the output embedding
+        frequency_embedding_size (`int`, optional): Size of the intermediate frequency embedding
+    """
+    def __init__(self, hidden_size, frequency_embedding_size=256):
+        super().__init__()
+        self.mlp = nn.Sequential(
+            nn.Linear(frequency_embedding_size, hidden_size, bias=False),
+            # nn.SiLU(),
+            ACT2FN['silu'],
+            nn.Linear(hidden_size, hidden_size, bias=False),
+        )
+        self.frequency_embedding_size = frequency_embedding_size
+
+    @staticmethod
+    def timestep_embedding(t, dim, max_period=10000):
+        """
+        Create sinusoidal timestep embeddings.
+        
+        Args:
+            t (`torch.Tensor`): A 1-D Tensor of N indices, one per batch element.
+                            These may be fractional.
+            dim (`int`): The dimension of the output.
+            max_period (`int`, optional): Controls the minimum frequency of the embeddings.
+            
+        Returns:
+            `torch.Tensor`: An [N, D] Tensor of positional embeddings.
+        """
+        half = dim // 2
+        freqs = torch.exp(
+            -math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half
+        ).to(t.device)
+        args = t[:, None].float() * freqs[None]
+        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
+        if dim % 2:
+            embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
+        return embedding.to(t.dtype)
+
+    def forward(self, t):
+        t_freq = self.timestep_embedding(t, self.frequency_embedding_size)
+        t_emb = self.mlp(t_freq)
+        return t_emb
+
+
+class FeedForwardNetwork(nn.Module):
+    """
+    Standard feed-forward network with SwiGLU activation.
+    
+    Args:
+        embed_dim (`int`): Input dimension
+        ffn_dim (`int`): Hidden dimension
+    """
+    def __init__(
+        self,
+        embed_dim,
+        ffn_dim,
+    ):
+        super().__init__()
+        self.embed_dim = embed_dim
+        self.gate_proj = nn.Linear(self.embed_dim, ffn_dim, bias=False)
+        self.up_proj = nn.Linear(self.embed_dim, ffn_dim, bias=False)
+        self.down_proj = nn.Linear(ffn_dim, self.embed_dim, bias=False)
+        self.act_fn = ACT2FN['silu']  # Using SiLU as the activation function
+
+    def forward(self, x):
+        gate = self.gate_proj(x)
+        up = self.up_proj(x)
+        
+        # SwiGLU activation
+        # gate = F.silu(gate)
+        gate = self.act_fn(gate)
+        return self.down_proj(gate * up)
+
+    
+class HeadLayer(nn.Module):
+    """
+    A layer in the diffusion head.
+    
+    Args:
+        embed_dim (`int`): Input dimension
+        ffn_dim (`int`): Hidden dimension
+        cond_dim (`int`): Condition embedding dimension
+        norm_eps (`float`, optional): Epsilon for normalization
+    """
+    def __init__(
+        self,
+        embed_dim,
+        ffn_dim,
+        cond_dim,
+        norm_eps=1e-5,
+    ):
+        super().__init__()
+        self.embed_dim = embed_dim
+        self.cond_dim = cond_dim
+        self.ffn_dim = ffn_dim
+        self.ffn = FeedForwardNetwork(
+            self.embed_dim,
+            self.ffn_dim,
+        )
+        self.norm = RMSNorm(self.embed_dim, eps=norm_eps)
+        self.adaLN_modulation = nn.Sequential(
+            # nn.SiLU(),
+            ACT2FN['silu'],
+            nn.Linear(cond_dim, 3 * self.embed_dim, bias=False)
+        )
+
+    def forward(self, x, c):
+        shift_ffn, scale_ffn, gate_ffn = self.adaLN_modulation(c).chunk(3, dim=-1)
+        x = x + gate_ffn * self.ffn(modulate(self.norm(x), shift_ffn, scale_ffn))
+        return x
+
+
+class FinalLayer(nn.Module):
+    """
+    Final layer in the diffusion head.
+    
+    Args:
+        hidden_size (`int`): Input dimension
+        output_size (`int`): Output dimension
+        cond_size (`int`): Condition embedding dimension
+        norm_eps (`float`, optional): Epsilon for normalization
+    """
+    def __init__(self, hidden_size, output_size, cond_size, norm_eps=1e-5):
+        super().__init__()
+        self.norm_final = RMSNorm(hidden_size, eps=norm_eps, elementwise_affine=False)
+        self.linear = nn.Linear(hidden_size, output_size, bias=False)
+        self.adaLN_modulation = nn.Sequential(
+            # nn.SiLU(),
+            ACT2FN['silu'],
+            nn.Linear(cond_size, 2 * hidden_size, bias=False)
+        )
+
+    def forward(self, x, c):
+        shift, scale = self.adaLN_modulation(c).chunk(2, dim=-1)
+        x = modulate(self.norm_final(x), shift, scale)
+        x = self.linear(x)
+        return x
+
+
+class VibeVoiceDiffusionHead(PreTrainedModel):
+    """
+    Diffusion head model for vibevoice.
+    
+    Args:
+        config (`VibeVoiceDiffusionHeadConfig`): Model configuration
+        latent_size (`int`, optional): Size of the latent space. If not provided, uses `config.latent_size`.
+    """
+    config_class = VibeVoiceDiffusionHeadConfig
+    supports_gradient_checkpointing = True
+    _supports_flash_attn_2 = True  
+    _supports_sdpa = True  
+    
+    def __init__(
+        self,
+        config,
+    ):
+        super().__init__(config)
+        self.config = config
+        self.cond_dim = config.hidden_size
+        latent_size = config.latent_size
+        
+        self.noisy_images_proj = nn.Linear(latent_size, config.hidden_size, bias=False)
+        self.cond_proj = nn.Linear(config.hidden_size, self.cond_dim, bias=False)
+        self.t_embedder = TimestepEmbedder(self.cond_dim)
+        
+        ffn_dim = int(config.hidden_size * config.head_ffn_ratio)
+        
+        # Create the intermediate layers
+        self.layers = nn.ModuleList([
+            HeadLayer(
+                embed_dim=config.hidden_size,
+                ffn_dim=ffn_dim,
+                cond_dim=self.cond_dim,
+                norm_eps=config.rms_norm_eps
+            )
+            for _ in range(config.head_layers)
+        ])
+        
+        # Final layer for output
+        self.final_layer = FinalLayer(
+            hidden_size=config.hidden_size, 
+            output_size=latent_size,
+            cond_size=self.cond_dim,
+            norm_eps=config.rms_norm_eps
+        )
+        
+        self.initialize_weights()
+
+    def initialize_weights(self):
+        """Initialize the weights of the model."""
+        # Initialize timestep embedder
+        nn.init.normal_(self.t_embedder.mlp[0].weight, std=0.02)
+        nn.init.normal_(self.t_embedder.mlp[2].weight, std=0.02)
+
+        # Zero-out adaLN modulation layers
+        for layer in self.layers:
+            nn.init.constant_(layer.adaLN_modulation[-1].weight, 0)
+
+        # Zero-out output layers
+        nn.init.constant_(self.final_layer.adaLN_modulation[-1].weight, 0)
+        nn.init.constant_(self.final_layer.linear.weight, 0)
+
+    def forward(
+        self,
+        noisy_images,
+        timesteps,
+        condition,
+    ):
+        """
+        Forward pass of the prediction head.
+        
+        Args:
+            noisy_images (`torch.Tensor`): Noisy images/latents to denoise
+            timesteps (`torch.Tensor`): Timesteps for diffusion
+            condition (`torch.Tensor`): Conditioning information
+            
+        Returns:
+            `torch.Tensor`: The predicted noise/velocity
+        """
+        x = self.noisy_images_proj(noisy_images)
+        t = self.t_embedder(timesteps)
+        condition = self.cond_proj(condition)
+        c = condition + t
+        
+        for layer in self.layers:
+            x = layer(x, c)
+            
+        x = self.final_layer(x, c)
+        return x
+
+
+AutoModel.register(VibeVoiceDiffusionHeadConfig, VibeVoiceDiffusionHead)
+
+__all__ = [
+    "VibeVoiceDiffusionHead",
+]

vvembed/modular/modular_vibevoice_text_tokenizer.py

@@ -0,0 +1,214 @@
+"""Tokenization classes for vibevoice."""
+
+from typing import List, Optional, Union
+
+from transformers.utils import logging
+from transformers.models.qwen2.tokenization_qwen2 import Qwen2Tokenizer
+from transformers.models.qwen2.tokenization_qwen2_fast import Qwen2TokenizerFast
+
+logger = logging.get_logger(__name__)
+
+
+class VibeVoiceTextTokenizer(Qwen2Tokenizer):
+    """
+    Construct a VibeVoice tokenizer. Based on the Qwen2 tokenizer with additional special tokens for speech.
+    
+    Args:
+        vocab_file (`str`):
+            Path to the vocabulary file.
+        merges_file (`str`):
+            Path to the merges file.
+        errors (`str`, *optional*, defaults to `"replace"`):
+            Paradigm to follow when decoding bytes to UTF-8.
+        unk_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The unknown token.
+        bos_token (`str`, *optional*):
+            The beginning of sequence token. Not used for vibevoice.
+        eos_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The end of sequence token.
+        pad_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The token used for padding.
+        add_special_tokens (`bool`, *optional*, defaults to `True`):
+            Whether or not to add special tokens when encoding.
+    """
+
+    model_input_names = ["input_ids", "attention_mask"]
+
+    def __init__(
+        self,
+        vocab_file,
+        merges_file,
+        errors="replace",
+        unk_token="<|endoftext|>",
+        bos_token=None,
+        eos_token="<|endoftext|>",
+        pad_token="<|endoftext|>",
+        add_prefix_space=False,
+        add_special_tokens=True,
+        **kwargs,
+    ):
+        super().__init__(
+            vocab_file=vocab_file,
+            merges_file=merges_file,
+            errors=errors,
+            unk_token=unk_token,
+            bos_token=bos_token,
+            eos_token=eos_token,
+            pad_token=pad_token,
+            add_prefix_space=add_prefix_space,
+            add_special_tokens=add_special_tokens,
+            **kwargs,
+        )
+        
+        # Add VibeVoice-specific special tokens
+        self._add_vibevoice_special_tokens()
+        
+    def _add_vibevoice_special_tokens(self):
+        """Add VibeVoice-specific special tokens."""
+        special_tokens = {
+            "additional_special_tokens": [
+                "<|vision_start|>",  # Speech start (reusing vision tokens)
+                "<|vision_end|>",  # Speech end
+                "<|vision_pad|>",  # Speech diffusion pad
+            ]
+        }
+        num_added = self.add_special_tokens(special_tokens)
+        
+        # Cache special token IDs
+        self._speech_start_id = self.convert_tokens_to_ids("<|vision_start|>")
+        self._speech_end_id = self.convert_tokens_to_ids("<|vision_end|>")
+        self._speech_diffusion_id = self.convert_tokens_to_ids("<|vision_pad|>")
+        
+        self._eos_id = self.convert_tokens_to_ids('<|endoftext|>')
+
+        return num_added
+    
+    @property
+    def eos_id(self) -> int:
+        """Id of the end of sequence token."""
+        return self._eos_id
+    
+    @property
+    def speech_start_id(self) -> int:
+        """Id of the speech start token."""
+        return self._speech_start_id
+    
+    @property
+    def speech_end_id(self) -> int:
+        """Id of the speech end token."""
+        return self._speech_end_id
+    
+    @property
+    def speech_diffusion_id(self) -> int:
+        """Id of the speech diffusion token."""
+        return self._speech_diffusion_id
+    
+    @property
+    def pad_id(self) -> int:
+        """Id used for padding (returns -100 for loss masking)."""
+        return -100
+
+
+class VibeVoiceTextTokenizerFast(Qwen2TokenizerFast):
+    """
+    Construct a "fast" VibeVoice tokenizer (backed by HuggingFace's *tokenizers* library).
+    Based on the Qwen2 tokenizer with additional special tokens for speech.
+    
+    Args:
+        vocab_file (`str`, *optional*):
+            Path to the vocabulary file.
+        merges_file (`str`, *optional*):
+            Path to the merges file.
+        tokenizer_file (`str`, *optional*):
+            Path to [tokenizers](https://github.com/huggingface/tokenizers) file.
+        unk_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The unknown token.
+        bos_token (`str`, *optional*):
+            The beginning of sequence token. Not used for vibevoice.
+        eos_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The end of sequence token.
+        pad_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The token used for padding.
+    """
+
+    model_input_names = ["input_ids", "attention_mask"]
+
+    def __init__(
+        self,
+        vocab_file=None,
+        merges_file=None,
+        tokenizer_file=None,
+        unk_token="<|endoftext|>",
+        bos_token=None,
+        eos_token="<|endoftext|>",
+        pad_token="<|endoftext|>",
+        add_prefix_space=False,
+        **kwargs,
+    ):
+        super().__init__(
+            vocab_file=vocab_file,
+            merges_file=merges_file,
+            tokenizer_file=tokenizer_file,
+            unk_token=unk_token,
+            bos_token=bos_token,
+            eos_token=eos_token,
+            pad_token=pad_token,
+            add_prefix_space=add_prefix_space,
+            **kwargs,
+        )
+        
+        # Add VibeVoice-specific special tokens
+        self._add_vibevoice_special_tokens()
+        
+    def _add_vibevoice_special_tokens(self):
+        """Add VibeVoice-specific special tokens."""
+        special_tokens = {
+            "additional_special_tokens": [
+                "<|vision_start|>",  # Speech start (reusing vision tokens)
+                "<|vision_end|>",  # Speech end
+                "<|vision_pad|>",  # Speech diffusion pad
+            ]
+        }
+        num_added = self.add_special_tokens(special_tokens)
+        
+        # Cache special token IDs
+        self._speech_start_id = self.convert_tokens_to_ids("<|vision_start|>")
+        self._speech_end_id = self.convert_tokens_to_ids("<|vision_end|>")
+        self._speech_diffusion_id = self.convert_tokens_to_ids("<|vision_pad|>")
+
+        # self._eos_id = self.convert_tokens_to_ids('<|endoftext|>')
+        self._eos_id = self.eos_token_id # qwen2 / qwen3
+        self._pad_id = self.convert_tokens_to_ids('<|image_pad|>')
+        
+        return num_added
+    
+    @property
+    def eos_id(self) -> int:
+        """Id of the end of sequence token."""
+        return self._eos_id
+    
+    @property
+    def speech_start_id(self) -> int:
+        """Id of the speech start token."""
+        return self._speech_start_id
+    
+    @property
+    def speech_end_id(self) -> int:
+        """Id of the speech end token."""
+        return self._speech_end_id
+    
+    @property
+    def speech_diffusion_id(self) -> int:
+        """Id of the speech diffusion token."""
+        return self._speech_diffusion_id
+    
+    @property
+    def pad_id(self) -> int:
+        """Id used for padding (returns -100 for loss masking)."""
+        return self._pad_id
+
+
+__all__ = [
+    "VibeVoiceTextTokenizer", 
+    "VibeVoiceTextTokenizerFast", 
+]

vvembed/modular/modular_vibevoice_tokenizer.py

@@ -0,0 +1,1195 @@
+import math
+import typing as tp
+from functools import partial
+from dataclasses import dataclass, field
+from typing import Dict, List, Optional, Tuple, Union
+import copy
+
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from transformers.models.auto import AutoModel
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+from transformers.modeling_utils import PreTrainedModel
+from transformers.activations import ACT2FN
+
+from .configuration_vibevoice import VibeVoiceAcousticTokenizerConfig, VibeVoiceSemanticTokenizerConfig
+
+logger = logging.get_logger(__name__)
+
+import os
+# Try to import APEX FusedRMSNorm
+try:
+    from apex.normalization.fused_layer_norm import fused_rms_norm_affine
+    APEX_AVAILABLE = True
+    logger.info("APEX FusedRMSNorm is available and will be used for optimization")
+    if int(os.getenv("OPTIMIZE_FOR_SPEED", "0")) == 0:
+        APEX_AVAILABLE = False
+        logger.warning("APEX FusedRMSNorm is disabled by environment variable OPTIMIZE_FOR_SPEED=0")
+except ImportError:
+    APEX_AVAILABLE = False
+    logger.warning("APEX FusedRMSNorm not available, using native implementation")
+# APEX_AVAILABLE=False
+
+# Normalization modules
+class ConvLayerNorm(nn.LayerNorm):
+    """
+    Convolution-friendly LayerNorm that moves channels to last dimensions
+    before running the normalization and moves them back to original position right after.
+    """
+    def __init__(self, normalized_shape: tp.Union[int, tp.List[int], torch.Size], **kwargs):
+        super().__init__(normalized_shape, **kwargs)
+
+    def forward(self, x):
+        x = x.transpose(1, 2)  # b ... t -> b t ...
+        x = nn.functional.layer_norm(x.float(), self.normalized_shape, self.weight.float(), self.bias.float(), self.eps).type_as(x) 
+        x = x.transpose(1, 2)  # b t ... -> b ... t
+        return x
+    
+class RMSNorm(nn.Module):
+    def __init__(self, dim: int, eps: float = 1e-5, elementwise_affine=True, weight_shape=None):
+        super().__init__()
+        self.dim = dim
+        self.eps = eps
+        self.elementwise_affine = elementwise_affine
+        if self.elementwise_affine:
+            weight_shape = (dim,) if weight_shape is None else weight_shape
+            self.weight = nn.Parameter(torch.ones(weight_shape))
+        else:
+            self.register_parameter('weight', None)
+
+    def _norm(self, x):
+        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)
+
+    def forward(self, x):
+        output = self._norm(x.float()).type_as(x)
+        if self.weight is not None:
+            output = output * self.weight
+        return output
+
+    def extra_repr(self) -> str:
+        return f'dim={self.dim}, eps={self.eps}, elementwise_affine={self.elementwise_affine}'
+
+class ConvRMSNorm(RMSNorm):
+    def __init__(self, dim: int, eps: float = 1e-5, elementwise_affine=True, weight_shape=None):
+        super().__init__(dim, eps, elementwise_affine, weight_shape)
+
+    def forward(self, x):
+        x = x.transpose(1, 2)  # b ... t -> b t ...
+        if (not APEX_AVAILABLE) or (not self.elementwise_affine):
+            # Fallback to native implementation
+            output = self._norm(x.float()).type_as(x)
+            if self.weight is not None:
+                output = output * self.weight
+        else:
+            output = fused_rms_norm_affine(x, self.weight, self.weight.shape, self.eps)
+        output = output.transpose(1, 2)  # b t ... -> b ... t
+        return output
+
+# Convolutional layers and utilities
+CONV_NORMALIZATIONS = frozenset(['none', 'weight_norm', 'spectral_norm',
+                                'time_layer_norm', 'layer_norm', 'time_group_norm'])
+
+
+def apply_parametrization_norm(module: nn.Module, norm: str = 'none') -> nn.Module:
+    assert norm in CONV_NORMALIZATIONS
+    if norm == 'weight_norm':
+        return nn.utils.weight_norm(module)
+    elif norm == 'spectral_norm':
+        return nn.utils.spectral_norm(module)
+    else:
+        # We already check was in CONV_NORMALIZATION, so any other choice
+        # doesn't need reparametrization.
+        return module
+
+
+def get_norm_module(module: nn.Module, causal: bool = False, norm: str = 'none', **norm_kwargs) -> nn.Module:
+    """Return the proper normalization module. If causal is True, this will ensure the returned
+    module is causal, or return an error if the normalization doesn't support causal evaluation.
+    """
+    assert norm in CONV_NORMALIZATIONS
+    if norm == 'layer_norm':
+        assert isinstance(module, nn.modules.conv._ConvNd)
+        return ConvLayerNorm(module.out_channels, **norm_kwargs)
+    elif norm == 'time_group_norm':
+        if causal:
+            raise ValueError("GroupNorm doesn't support causal evaluation.")
+        assert isinstance(module, nn.modules.conv._ConvNd)
+        return nn.GroupNorm(1, module.out_channels, **norm_kwargs)
+    else:
+        return nn.Identity()
+
+
+def get_extra_padding_for_conv1d(x: torch.Tensor, kernel_size: int, stride: int,
+                                padding_total: int = 0) -> int:
+    """Calculate extra padding needed for convolution to have the same output length"""
+    length = x.shape[-1]
+    n_frames = (length - kernel_size + padding_total) / stride + 1
+    ideal_length = (math.ceil(n_frames) - 1) * stride + (kernel_size - padding_total)
+    return ideal_length - length
+
+
+def pad1d(x: torch.Tensor, paddings: tp.Tuple[int, int], mode: str = 'zero', value: float = 0.):
+    """Pad 1D input with handling for small inputs in reflect mode"""
+    length = x.shape[-1]
+    padding_left, padding_right = paddings
+    assert padding_left >= 0 and padding_right >= 0, (padding_left, padding_right)
+    if mode == 'reflect':
+        max_pad = max(padding_left, padding_right)
+        extra_pad = 0
+        if length <= max_pad:
+            extra_pad = max_pad - length + 1
+            x = F.pad(x, (0, extra_pad))
+        padded = F.pad(x, paddings, mode, value)
+        end = padded.shape[-1] - extra_pad
+        return padded[..., :end]
+    else:
+        return F.pad(x, paddings, mode, value)
+
+
+def unpad1d(x: torch.Tensor, paddings: tp.Tuple[int, int]):
+    """Remove padding from x, handling properly zero padding. Only for 1d!"""
+    padding_left, padding_right = paddings
+    assert padding_left >= 0 and padding_right >= 0, (padding_left, padding_right)
+    assert (padding_left + padding_right) <= x.shape[-1]
+    end = x.shape[-1] - padding_right
+    return x[..., padding_left: end]
+
+
+class NormConv1d(nn.Module):
+    """Wrapper around Conv1d and normalization applied to this conv"""
+    def __init__(self, *args, causal: bool = False, norm: str = 'none',
+                norm_kwargs: tp.Dict[str, tp.Any] = {}, **kwargs):
+        super().__init__()
+        self.conv = apply_parametrization_norm(nn.Conv1d(*args, **kwargs), norm)
+        self.norm = get_norm_module(self.conv, causal, norm, **norm_kwargs)
+        self.norm_type = norm
+
+    def forward(self, x):
+        x = self.conv(x)
+        x = self.norm(x)
+        return x
+
+
+class NormConvTranspose1d(nn.Module):
+    """Wrapper around ConvTranspose1d and normalization applied to this conv"""
+    def __init__(self, *args, causal: bool = False, norm: str = 'none',
+                norm_kwargs: tp.Dict[str, tp.Any] = {}, **kwargs):
+        super().__init__()
+        self.convtr = apply_parametrization_norm(nn.ConvTranspose1d(*args, **kwargs), norm)
+        self.norm = get_norm_module(self.convtr, causal, norm, **norm_kwargs)
+        self.norm_type = norm
+
+    def forward(self, x):
+        x = self.convtr(x)
+        x = self.norm(x)
+        return x
+
+
+class VibeVoiceTokenizerStreamingCache:
+    """Cache for streaming convolution, similar to KV cache in attention"""
+    def __init__(self):
+        self.cache = {}  # Dict mapping (layer_id, sample_idx) to state tensor
+        
+    def get(self, layer_id: str, sample_indices: torch.Tensor) -> Optional[torch.Tensor]:
+        """Get cached states for given layer and sample indices"""
+        states = []
+        max_length = 0
+        
+        # First pass: collect states and find max length
+        for idx in sample_indices.tolist():
+            key = (layer_id, idx)
+            if key not in self.cache:
+                return None  # If any sample is missing, return None
+            state = self.cache[key]
+            states.append(state)
+            max_length = max(max_length, state.shape[-1])
+        
+        # Second pass: pad states to max length if needed
+        if len(states) > 0 and states[0].dim() >= 2:
+            padded_states = []
+            for state in states:
+                if state.shape[-1] < max_length:
+                    # Pad on the time dimension (last dimension)
+                    pad_size = max_length - state.shape[-1]
+                    # Pad with zeros on the LEFT to align the most recent samples
+                    padded_state = F.pad(state, (pad_size, 0), mode='constant', value=0)
+                    padded_states.append(padded_state)
+                else:
+                    padded_states.append(state)
+            return torch.stack(padded_states, dim=0)
+        else:
+            return torch.stack(states, dim=0)
+    
+    def set(self, layer_id: str, sample_indices: torch.Tensor, states: torch.Tensor):
+        """Set cached states for given layer and sample indices"""
+        for i, idx in enumerate(sample_indices.tolist()):
+            key = (layer_id, idx)
+            self.cache[key] = states[i].detach()
+
+    def set_to_zero(self, sample_indices: torch.Tensor):
+        """Set all cached states to zero for given sample indices"""
+        for key in list(self.cache.keys()):
+            layer_id, sample_idx = key
+            if sample_idx in sample_indices.tolist():
+                # Create zero tensor with same shape and dtype as cached tensor
+                cached_tensor = self.cache[key]
+                self.cache[key] = torch.zeros_like(cached_tensor)
+                
+    def clear(self, layer_id: Optional[str] = None, sample_indices: Optional[torch.Tensor] = None):
+        """Clear cache for specific layer/samples or everything"""
+        if layer_id is None and sample_indices is None:
+            self.cache.clear()
+        elif layer_id is not None and sample_indices is None:
+            # Clear all samples for a specific layer
+            keys_to_remove = [k for k in self.cache.keys() if k[0] == layer_id]
+            for k in keys_to_remove:
+                del self.cache[k]
+        elif layer_id is not None and sample_indices is not None:
+            # Clear specific samples for a specific layer
+            for idx in sample_indices.tolist():
+                key = (layer_id, idx)
+                self.cache.pop(key, None)
+
+class SConv1d(nn.Module):
+    """Conv1d with built-in handling of asymmetric or causal padding and normalization."""
+    def __init__(self, in_channels: int, out_channels: int,
+                kernel_size: int, stride: int = 1, dilation: int = 1,
+                groups: int = 1, bias: bool = True, causal: bool = False,
+                norm: str = 'none', norm_kwargs: tp.Dict[str, tp.Any] = {},
+                pad_mode: str = 'reflect'):
+        super().__init__()
+        self.conv = NormConv1d(in_channels, out_channels, kernel_size, stride,
+                            dilation=dilation, groups=groups, bias=bias, causal=causal,
+                            norm=norm, norm_kwargs=norm_kwargs)
+        self.causal = causal
+        self.pad_mode = pad_mode
+        
+        # Store configuration
+        self.kernel_size = kernel_size
+        self.dilation = dilation
+        self.stride = stride
+        self.in_channels = in_channels
+        self.out_channels = out_channels
+        
+        # For causal convolution, we need to maintain kernel_size - 1 samples as context
+        # need to check use which context_size is more suitable
+        # self.context_size = (kernel_size - 1) * dilation
+        self.context_size = (kernel_size - 1) * dilation - (stride - 1)
+        
+        # For non-streaming mode, calculate padding
+        self.padding_total = (kernel_size - 1) * dilation - (stride - 1)
+        
+        # Create a unique layer ID for cache management
+        self._layer_id = None
+                  
+    @property
+    def layer_id(self):
+        if self._layer_id is None:
+            self._layer_id = f"sconv1d_{id(self)}"
+        return self._layer_id
+        
+    def forward(self, x: torch.Tensor, 
+                cache: Optional[VibeVoiceTokenizerStreamingCache] = None,
+                sample_indices: Optional[torch.Tensor] = None,
+                use_cache: bool = False,
+                debug: bool = False) -> torch.Tensor:
+        """
+        Forward pass with optional streaming support via cache.
+        
+        Args:
+            x: Input tensor [batch_size, channels, time]
+            cache: VibeVoiceTokenizerStreamingCache object for maintaining states
+            sample_indices: Indices identifying each sample for cache management
+            use_cache: Whether to use cached states for streaming
+            debug: Whether to print debug information
+            
+        Returns:
+            Output tensor
+        """
+        B, C, T = x.shape
+        
+        # Non-streaming mode
+        if not use_cache or cache is None:
+            return self._forward_non_streaming(x, debug=debug)
+        
+        # Streaming mode
+        assert self.causal, "Streaming mode is only supported for causal convolutions"
+        assert sample_indices is not None, "sample_indices must be provided for streaming mode"
+        assert len(sample_indices) == B, "sample_indices must match batch size"
+        
+        return self._forward_streaming(x, cache, sample_indices, debug)
+    
+    def _forward_streaming(self, x: torch.Tensor, 
+                          cache: VibeVoiceTokenizerStreamingCache,
+                          sample_indices: torch.Tensor,
+                          debug: bool = False) -> torch.Tensor:
+        """Streaming forward pass with cache operations kept separate from compiled code"""
+        B, C, T = x.shape
+        
+        # Cache operations (not compiled)
+        cached_states = cache.get(self.layer_id, sample_indices)
+        
+        if cached_states is None:
+            # First chunk - initialize with zeros for context
+            if self.context_size > 0:
+                cached_states = torch.zeros(B, C, self.context_size, device=x.device, dtype=x.dtype)
+                if debug:
+                    print(f"[DEBUG] Initialized cache with shape: {cached_states.shape}, context_size={self.context_size}")
+            else:
+                cached_states = torch.zeros(B, C, 0, device=x.device, dtype=x.dtype)
+                if debug:
+                    print(f"[DEBUG] No context needed (kernel_size=stride)")
+        
+        # Concatenate cached states with input
+        if cached_states.shape[2] > 0:
+            input_with_context = torch.cat([cached_states, x], dim=2)
+        else:
+            input_with_context = x
+            
+        if debug:
+            print(f"[DEBUG] Input shape: {x.shape}, Cache shape: {cached_states.shape}, Combined: {input_with_context.shape}")
+        
+        # Apply convolution directly - no extra padding in streaming mode
+        # The conv layer will handle its own padding internally
+        output = self.conv(input_with_context)
+
+        if debug:
+            print(f"[DEBUG] Output shape: {output.shape}")
+        
+        # Update cache for next chunk
+        if self.context_size > 0:
+            # Calculate how many samples to keep
+            total_input_length = input_with_context.shape[2]
+            
+            # Keep the last context_size samples
+            if total_input_length >= self.context_size:
+                new_cache_start = total_input_length - self.context_size
+                new_cache = input_with_context[:, :, new_cache_start:]
+            else:
+                # If we have less than context_size samples, keep everything
+                new_cache = input_with_context
+                
+            if debug:
+                print(f"[DEBUG] New cache shape: {new_cache.shape}")
+                
+            cache.set(self.layer_id, sample_indices, new_cache)
+        
+        return output
+    
+    def _forward_non_streaming(self, x: torch.Tensor, debug: bool = False) -> torch.Tensor:
+        """Standard forward pass without streaming"""
+        B, C, T = x.shape
+        kernel_size = self.kernel_size
+        stride = self.stride
+        dilation = self.dilation
+        padding_total = self.padding_total
+        
+        # Compute extra padding for stride alignment
+        extra_padding = get_extra_padding_for_conv1d(x, kernel_size, stride, padding_total)
+        
+        if debug:
+            print(f"[DEBUG NON-STREAMING] Input shape: {x.shape}, padding_total={padding_total}, extra_padding={extra_padding}")
+        
+        if self.causal:
+            # Left padding for causal
+            if self.pad_mode == 'constant':
+                x = pad1d(x, (padding_total, extra_padding), mode=self.pad_mode, value=0)
+            else:
+                x = pad1d(x, (padding_total, extra_padding), mode=self.pad_mode)
+        else:
+            # Symmetric padding for non-causal
+            padding_right = padding_total // 2
+            padding_left = padding_total - padding_right
+            x = pad1d(x, (padding_left, padding_right + extra_padding), mode=self.pad_mode)
+        
+        if debug:
+            print(f"[DEBUG NON-STREAMING] After padding: {x.shape}")
+            
+        output = self.conv(x)
+        
+        if debug:
+            print(f"[DEBUG NON-STREAMING] Output shape: {output.shape}")
+        
+        return output
+
+
+class SConvTranspose1d(nn.Module):
+    """ConvTranspose1d with built-in handling of asymmetric or causal padding and normalization."""
+    def __init__(self, in_channels: int, out_channels: int,
+                kernel_size: int, stride: int = 1, causal: bool = False,
+                norm: str = 'none', trim_right_ratio: float = 1.,
+                norm_kwargs: tp.Dict[str, tp.Any] = {}, bias: bool = True):
+        super().__init__()
+        self.convtr = NormConvTranspose1d(in_channels, out_channels, kernel_size, stride,
+                                        causal=causal, norm=norm, norm_kwargs=norm_kwargs, bias=bias)
+        self.causal = causal
+        self.trim_right_ratio = trim_right_ratio
+        assert self.causal or self.trim_right_ratio == 1., \
+            "`trim_right_ratio` != 1.0 only makes sense for causal convolutions"
+        assert self.trim_right_ratio >= 0. and self.trim_right_ratio <= 1.
+
+        # Store configuration
+        self.kernel_size = kernel_size
+        self.stride = stride
+        self.in_channels = in_channels
+        self.out_channels = out_channels
+        
+        # For transposed convolution, padding calculation is different
+        self.padding_total = kernel_size - stride
+        
+        # For streaming, we need to keep track of input history
+        # Transposed conv needs to see multiple input samples to produce correct output
+        self.context_size = kernel_size - 1
+        
+        # Create a unique layer ID for cache management
+        self._layer_id = None
+
+    @property
+    def layer_id(self):
+        if self._layer_id is None:
+            self._layer_id = f"sconvtr1d_{id(self)}"
+        return self._layer_id
+    
+    def forward(self, x: torch.Tensor,
+                cache: Optional[VibeVoiceTokenizerStreamingCache] = None,
+                sample_indices: Optional[torch.Tensor] = None,
+                use_cache: bool = False,
+                debug: bool = False) -> torch.Tensor:
+        """
+        Forward pass with optional streaming support via cache.
+        """
+        B, C, T = x.shape
+        
+        # Non-streaming mode
+        if not use_cache or cache is None:
+            return self._forward_non_streaming(x, debug=debug)
+        
+        # Streaming mode
+        assert sample_indices is not None, "sample_indices must be provided for streaming mode"
+        assert len(sample_indices) == B, "sample_indices must match batch size"
+        
+        return self._forward_streaming(x, cache, sample_indices, debug)
+    
+    def _forward_streaming(self, x: torch.Tensor,
+                          cache: VibeVoiceTokenizerStreamingCache,
+                          sample_indices: torch.Tensor,
+                          debug: bool = False) -> torch.Tensor:
+        """Streaming forward pass with cache operations kept separate from compiled code"""
+        B, C, T = x.shape
+        
+        # Cache operations (not compiled)
+        cached_input = cache.get(self.layer_id, sample_indices)
+        
+        if cached_input is None:
+            # First chunk - no history yet
+            cached_input = torch.zeros(B, C, 0, device=x.device, dtype=x.dtype)
+            if debug:
+                print(f"[DEBUG] Initialized empty cache for transposed conv")
+        
+        # Concatenate cached input with new input
+        full_input = torch.cat([cached_input, x], dim=2)
+        
+        if debug:
+            print(f"[DEBUG] Input shape: {x.shape}, Cache shape: {cached_input.shape}, Combined: {full_input.shape}")
+        
+        # First chunk or debug mode - use uncompiled version
+        full_output = self.convtr(full_input)
+        
+        if debug:
+            print(f"[DEBUG] Full transposed conv output shape: {full_output.shape}")
+        
+        # Calculate padding to remove
+        if self.causal:
+            padding_right = math.ceil(self.padding_total * self.trim_right_ratio)
+            padding_left = self.padding_total - padding_right
+        else:
+            padding_right = self.padding_total // 2
+            padding_left = self.padding_total - padding_right
+        
+        # Remove padding
+        if padding_left + padding_right > 0:
+            full_output = unpad1d(full_output, (padding_left, padding_right))
+        
+        if debug:
+            print(f"[DEBUG] After unpadding: {full_output.shape}")
+        
+        # Determine which part of the output corresponds to the new input
+        if cached_input.shape[2] == 0:
+            # First chunk - return all output
+            output = full_output
+        else:
+            # Subsequent chunks - return only the new output
+            expected_new_output = T * self.stride
+            
+            # Take the last expected_new_output samples
+            if full_output.shape[2] >= expected_new_output:
+                output = full_output[:, :, -expected_new_output:]
+            else:
+                output = full_output
+        
+        if debug:
+            print(f"[DEBUG] Final streaming output shape: {output.shape}")
+        
+        # Update cache
+        if full_input.shape[2] > self.context_size:
+            new_cache = full_input[:, :, -self.context_size:]
+        else:
+            new_cache = full_input
+        
+        if debug:
+            print(f"[DEBUG] New cache shape: {new_cache.shape}")
+            
+        cache.set(self.layer_id, sample_indices, new_cache)
+        
+        return output
+    
+    def _forward_non_streaming(self, x: torch.Tensor, debug: bool = False) -> torch.Tensor:
+        """Standard forward pass without streaming"""
+        if debug:
+            print(f"[DEBUG NON-STREAMING] Input shape: {x.shape}")
+        
+        # Apply transposed convolution
+        y = self.convtr(x)
+        
+        if debug:
+            print(f"[DEBUG NON-STREAMING] After transposed conv: {y.shape}")
+        
+        # Calculate and remove padding
+        if self.causal:
+            padding_right = math.ceil(self.padding_total * self.trim_right_ratio)
+            padding_left = self.padding_total - padding_right
+        else:
+            padding_right = self.padding_total // 2
+            padding_left = self.padding_total - padding_right
+        
+        if padding_left + padding_right > 0:
+            y = unpad1d(y, (padding_left, padding_right))
+        
+        if debug:
+            print(f"[DEBUG NON-STREAMING] Final output shape: {y.shape}")
+            
+        return y
+    
+# FFN 
+class FFN(nn.Module):
+    def __init__(
+        self,
+        embed_dim,
+        ffn_dim,
+        bias=False,
+    ):
+        super().__init__()
+        self.embed_dim = embed_dim
+        self.linear1 = nn.Linear(self.embed_dim, ffn_dim, bias=bias) 
+        self.gelu = ACT2FN["gelu"]
+        self.linear2 = nn.Linear(ffn_dim, self.embed_dim, bias=bias)
+
+    def forward(self, x):
+        x = self.linear1(x)
+        x = self.gelu(x)
+        x = self.linear2(x)
+        return x
+
+
+class Convlayer(nn.Module):
+    def __init__(
+            self, 
+            in_channels, 
+            out_channels, 
+            kernel_size, 
+            stride=1, 
+            dilation=1, 
+            groups=1, 
+            bias=True, 
+            pad_mode='zeros', 
+            norm='weight_norm', 
+            causal=True, 
+        ):
+        super().__init__()
+        self.conv = SConv1d(in_channels, out_channels, kernel_size, stride=stride, dilation=dilation, 
+                           groups=groups, bias=bias, pad_mode=pad_mode, norm=norm, causal=causal)
+
+    def forward(self, x):
+        return self.conv(x)
+
+class Block1D(nn.Module):
+    def __init__(self, dim, kernel_size=7, drop_path=0., mixer_layer='conv',  
+                layer_scale_init_value=1e-6, **kwargs):
+        super().__init__()
+        
+        if kwargs.get('layernorm', 'LN') == 'LN':
+            self.norm = ConvLayerNorm(dim, eps=kwargs.get('eps', 1e-6))
+            self.ffn_norm = ConvLayerNorm(dim, eps=kwargs.get('eps', 1e-6))               
+        elif kwargs.get('layernorm', 'RMSNorm') == 'RMSNorm':
+            self.norm = ConvRMSNorm(dim, eps=kwargs.get('eps', 1e-6))
+            self.ffn_norm = ConvRMSNorm(dim, eps=kwargs.get('eps', 1e-6))
+
+        if mixer_layer == 'conv':
+            self.mixer = Convlayer(dim, dim, groups=kwargs.get('groups', 1),
+                                kernel_size=kernel_size, 
+                                pad_mode=kwargs.get('pad_mode', 'reflect'), 
+                                norm=kwargs.get('norm', 'none'), 
+                                causal=kwargs.get('causal', True), 
+                                bias=kwargs.get('bias', True),
+                                )
+        elif mixer_layer == 'depthwise_conv':
+            self.mixer = Convlayer(dim, dim, groups=dim,
+                                kernel_size=kernel_size, 
+                                pad_mode=kwargs.get('pad_mode', 'reflect'), 
+                                norm=kwargs.get('norm', 'none'), 
+                                causal=kwargs.get('causal', True), 
+                                bias=kwargs.get('bias', True),
+                                )
+        else:
+            raise ValueError(f"Unsupported mixer layer: {mixer_layer}")
+        
+        self.ffn = FFN(
+            dim, 
+            kwargs.get('ffn_expansion', 4) * dim, 
+            bias=kwargs.get('bias', False),
+        )
+        self.drop_path = nn.Identity() if drop_path <= 0. else nn.modules.DropPath(drop_path)
+
+        if layer_scale_init_value > 0:
+            self.gamma = nn.Parameter(layer_scale_init_value * torch.ones((dim)), requires_grad=True)
+            self.ffn_gamma = nn.Parameter(layer_scale_init_value * torch.ones((dim)), requires_grad=True)
+        else:
+            self.gamma = None
+            self.ffn_gamma = None
+
+    def forward(self, x):
+        # mixer
+        residual = x
+        x = self.norm(x)
+        x = self.mixer(x)
+        if self.gamma is not None:
+            x = x * self.gamma.unsqueeze(-1)
+        x = residual + self.drop_path(x)
+
+        # ffn
+        residual = x
+        x = self.ffn_norm(x)
+        x = x.permute(0, 2, 1)
+        x = self.ffn(x)
+        x = x.permute(0, 2, 1)
+        if self.ffn_gamma is not None:
+            x = x * self.ffn_gamma.unsqueeze(-1)
+        x = residual + self.drop_path(x)
+
+        return x
+
+
+class TokenizerEncoder(nn.Module):
+    """
+    Encoder component for the VibeVoice tokenizer that converts audio to latent representations.
+    
+    Args:
+        config: Configuration object with model parameters
+    """
+    def __init__(self, config):
+        super().__init__()
+        
+        # Extract parameters from config
+        self.channels = config.channels
+        self.dimension = config.dimension
+        self.n_filters = config.n_filters
+        self.ratios = list(reversed(config.ratios))
+        self.depths = config.depths
+        self.n_residual_layers = getattr(config, "n_residual_layers", 1)
+        self.hop_length = np.prod(self.ratios)
+        self.causal = config.causal
+        
+        # Additional config parameters with defaults
+        kernel_size = getattr(config, "kernel_size", 7)
+        last_kernel_size = getattr(config, "last_kernel_size", 7)
+        norm = getattr(config, "norm", "none")
+        norm_params = getattr(config, "norm_params", {})
+        pad_mode = getattr(config, "pad_mode", "reflect")
+        bias = getattr(config, "bias", True)
+        layernorm = getattr(config, "layernorm", "LN")
+        layernorm_eps = getattr(config, "layernorm_eps", 1e-6)
+        layernorm_elementwise_affine = getattr(config, "layernorm_elementwise_affine", True)
+        drop_path_rate = getattr(config, "drop_path_rate", 0.0)
+        mixer_layer = getattr(config, "mixer_layer", "conv")
+        layer_scale_init_value = getattr(config, "layer_scale_init_value", 0)
+        disable_last_norm = getattr(config, "disable_last_norm", False)
+        
+        # determine the norm type based on layernorm
+        if layernorm == 'LN':
+            norm_type = ConvLayerNorm
+        elif layernorm == 'RMSNorm':
+            norm_type = partial(ConvRMSNorm, elementwise_affine=layernorm_elementwise_affine)
+        else:
+            raise ValueError(f"Unsupported norm type: {layernorm}")
+        
+        # stem and intermediate downsampling conv layers
+        stem = nn.Sequential(
+                SConv1d(self.channels, self.n_filters, kernel_size, norm=norm, norm_kwargs=norm_params, causal=self.causal, pad_mode=pad_mode, bias=bias),
+            )
+        
+        self.downsample_layers = nn.ModuleList()
+        self.downsample_layers.append(stem)
+        for i in range(len(self.ratios)):
+            in_ch = self.n_filters * (2 ** i)
+            out_ch = self.n_filters * (2 ** (i + 1))
+            downsample_layer = nn.Sequential(
+                SConv1d(in_ch, out_ch, kernel_size=self.ratios[i] * 2, stride=self.ratios[i], causal=self.causal, pad_mode=pad_mode, norm=norm, bias=bias)
+            )
+            self.downsample_layers.append(downsample_layer)
+
+        # configure the transformer blocks
+        layer_type = partial(
+            Block1D,
+            mixer_layer=mixer_layer,
+            layernorm=layernorm,
+            eps=layernorm_eps,
+            causal=self.causal,
+            pad_mode=pad_mode,
+            norm=norm,
+            bias=bias,
+            layer_scale_init_value=layer_scale_init_value,
+        )
+        
+        self.stages = nn.ModuleList()
+        dp_rates = [x.item() for x in torch.linspace(0, drop_path_rate, sum(self.depths))] 
+        cur = 0
+
+        for i in range(len(self.depths)):
+            in_ch = self.n_filters * (2 ** i)
+            stage = nn.Sequential(
+                *[layer_type(dim=in_ch, drop_path=dp_rates[cur + j]) for j in range(self.depths[i])]
+            )
+            self.stages.append(stage)
+            cur += self.depths[i]
+        
+        if not disable_last_norm:
+            self.norm = norm_type(in_ch, eps=layernorm_eps)
+        else:
+            self.norm = nn.Identity()
+        self.head = SConv1d(in_ch, self.dimension, kernel_size=last_kernel_size, causal=self.causal, pad_mode=pad_mode, norm=norm, bias=bias)
+
+    def forward_features(self, x, cache=None, sample_indices=None, use_cache=False, debug=False):
+        for i in range(len(self.depths)):
+            # Apply downsampling
+            for layer in self.downsample_layers[i]:
+                if isinstance(layer, SConv1d):
+                    x = layer(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+                else:
+                    x = layer(x)
+            
+            # Apply stage (Block1D contains Convlayer which contains SConv1d)
+            for block in self.stages[i]:
+                if hasattr(block, 'mixer') and hasattr(block.mixer, 'conv') and isinstance(block.mixer.conv, SConv1d):
+                    # Block1D forward with cache support
+                    residual = x
+                    x = block.norm(x)
+                    x = block.mixer.conv(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+                    if block.gamma is not None:
+                        x = x * block.gamma.unsqueeze(-1)
+                    x = residual + x
+                    
+                    # FFN part
+                    residual = x
+                    x = block.ffn_norm(x)
+                    x = x.permute(0, 2, 1)
+                    x = block.ffn(x)
+                    x = x.permute(0, 2, 1)
+                    if block.ffn_gamma is not None:
+                        x = x * block.ffn_gamma.unsqueeze(-1)
+                    x = residual + x
+                else:
+                    x = block(x)
+
+        return self.norm(x)
+
+    def forward(self, x, cache=None, sample_indices=None, use_cache=False, debug=False):
+        x = self.forward_features(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        x = self.head(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        return x
+
+
+class TokenizerDecoder(nn.Module):
+    """
+    Decoder component for the VibeVoice tokenizer that converts latent representations back to audio.
+    
+    Args:
+        config: Configuration object with model parameters
+    """
+    def __init__(self, config):
+        super().__init__()
+        
+        # Extract parameters from config
+        self.dimension = config.dimension
+        self.channels = config.channels
+        self.n_filters = config.n_filters
+        self.ratios = config.ratios
+        
+        # IMPORTANT CHANGE: Don't reverse depths again since they're already reversed in VibeVoiceAcousticTokenizerModel
+        self.depths = config.depths  # Changed from list(reversed(config.depths))
+        
+        self.n_residual_layers = getattr(config, "n_residual_layers", 1)
+        self.hop_length = np.prod(self.ratios)
+        self.causal = config.causal
+        
+        # Additional config parameters with defaults
+        kernel_size = getattr(config, "kernel_size", 7)
+        last_kernel_size = getattr(config, "last_kernel_size", 7)
+        norm = getattr(config, "norm", "none")
+        norm_params = getattr(config, "norm_params", {})
+        pad_mode = getattr(config, "pad_mode", "reflect")
+        bias = getattr(config, "bias", True)
+        layernorm = getattr(config, "layernorm", "LN")
+        layernorm_eps = getattr(config, "layernorm_eps", 1e-6)
+        trim_right_ratio = getattr(config, "trim_right_ratio", 1.0)
+        layernorm_elementwise_affine = getattr(config, "layernorm_elementwise_affine", True)
+        drop_path_rate = getattr(config, "drop_path_rate", 0.0)
+        mixer_layer = getattr(config, "mixer_layer", "conv")
+        layer_scale_init_value = getattr(config, "layer_scale_init_value", 0)
+        disable_last_norm = getattr(config, "disable_last_norm", False)
+
+        # determine the norm type based on layernorm
+        if layernorm == 'LN':
+            norm_type = ConvLayerNorm
+        elif layernorm == 'RMSNorm':
+            norm_type = partial(ConvRMSNorm, elementwise_affine=layernorm_elementwise_affine)
+        else:
+            raise ValueError(f"Unsupported norm type: {layernorm}")
+        
+        # stem and upsampling layers
+        stem = nn.Sequential(
+                SConv1d(self.dimension, self.n_filters * 2 ** (len(self.depths) - 1), kernel_size, norm=norm, 
+                        norm_kwargs=norm_params, causal=self.causal, pad_mode=pad_mode, bias=bias),
+            )
+        
+        self.upsample_layers = nn.ModuleList()
+        self.upsample_layers.append(stem)
+        for i in range(len(self.ratios)):
+            in_ch = self.n_filters * (2 ** (len(self.depths) - 1 - i))
+            out_ch = self.n_filters * (2 ** (len(self.depths) - 1 - i - 1))
+            upsample_layer = nn.Sequential(
+                SConvTranspose1d(in_ch, out_ch,
+                                kernel_size=self.ratios[i] * 2, stride=self.ratios[i],
+                                norm=norm, norm_kwargs=norm_params, bias=bias,
+                                causal=self.causal, trim_right_ratio=trim_right_ratio),
+            )
+            self.upsample_layers.append(upsample_layer)
+
+        # configure transformer blocks
+        layer_type = partial(
+            Block1D,
+            mixer_layer=mixer_layer,
+            layernorm=layernorm,
+            eps=layernorm_eps,
+            causal=self.causal,
+            pad_mode=pad_mode,
+            norm=norm,
+            bias=bias,
+            layer_scale_init_value=layer_scale_init_value,
+        )
+
+        self.stages = nn.ModuleList()
+        dp_rates = [x.item() for x in torch.linspace(0, drop_path_rate, sum(self.depths))] 
+        cur = 0
+        
+        # Create stages in the same order as the original model
+        for i in range(len(self.depths)):
+            in_ch = self.n_filters * (2 ** (len(self.depths) - 1 - i))
+            stage = nn.Sequential(
+                *[layer_type(dim=in_ch, drop_path=dp_rates[cur + j]) for j in range(self.depths[i])]
+            )
+            self.stages.append(stage)
+            cur += self.depths[i]
+
+        if not disable_last_norm:
+            self.norm = norm_type(in_ch, eps=layernorm_eps)
+        else:
+            self.norm = nn.Identity()
+        self.head = SConv1d(in_ch, self.channels, kernel_size=last_kernel_size, causal=self.causal, pad_mode=pad_mode, norm=norm, bias=bias)
+
+    def forward_features(self, x, cache=None, sample_indices=None, use_cache=False, debug=False):
+        for i in range(len(self.depths)):
+            # Apply upsampling
+            for layer in self.upsample_layers[i]:
+                if isinstance(layer, (SConv1d, SConvTranspose1d)):
+                    x = layer(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+                else:
+                    x = layer(x)
+            
+            # Apply stage (Block1D contains Convlayer which contains SConv1d)
+            for block in self.stages[i]:
+                if hasattr(block, 'mixer') and hasattr(block.mixer, 'conv') and isinstance(block.mixer.conv, SConv1d):
+                    # Block1D forward with cache support
+                    residual = x
+                    x = block.norm(x)
+                    x = block.mixer.conv(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+                    if block.gamma is not None:
+                        x = x * block.gamma.unsqueeze(-1)
+                    x = residual + x
+                    
+                    # FFN part
+                    residual = x
+                    x = block.ffn_norm(x)
+                    x = x.permute(0, 2, 1)
+                    x = block.ffn(x)
+                    x = x.permute(0, 2, 1)
+                    if block.ffn_gamma is not None:
+                        x = x * block.ffn_gamma.unsqueeze(-1)
+                    x = residual + x
+                else:
+                    x = block(x)
+
+        return self.norm(x)
+    
+    def forward(self, x, cache=None, sample_indices=None, use_cache=False, debug=False):
+        x = self.forward_features(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        x = self.head(x, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        return x
+    
+
+@dataclass
+class VibeVoiceTokenizerEncoderOutput:
+    """
+    Output of VibeVoice tokenizer encoder, representing a Gaussian distribution with fixed variance.
+    
+    Args:
+        mean (`torch.FloatTensor`): The mean parameters of the distribution.
+        std (`float` or `torch.FloatTensor`): Fixed standard deviation value.
+    """
+    mean: torch.Tensor
+    std: Optional[Union[float, torch.Tensor]] = None
+    
+    def sample(self, dist_type='fix'):
+        """
+        Sample from the distribution.
+        
+        Args:
+            dist_type (`str`): Sampling method, either 'fix' or 'gaussian'.
+                
+        Returns:
+            `torch.FloatTensor`: Sampled values.
+            `torch.FloatTensor` (optional): Standard deviation used (only when dist_type='gaussian').
+        """
+        if dist_type == 'fix':
+            x = self.mean + self.std * torch.randn_like(self.mean)
+            return x, self.std
+        elif dist_type == 'gaussian':
+            batch_size = self.mean.size(0)
+            value = self.std / 0.8
+            std = torch.randn(batch_size, device=self.mean.device, dtype=self.mean.dtype) * value
+
+            while std.dim() < self.mean.dim():
+                std = std.unsqueeze(-1)
+
+            x = self.mean + std * torch.randn_like(self.mean)
+            return x, std
+        else:
+            return self.mean, self.std
+
+    def kl(self):
+        """Compute KL divergence between this distribution and a standard normal."""
+        target = torch.zeros_like(self.mean)
+        return F.mse_loss(self.mean, target, reduction='none')
+
+    def mode(self):
+        """Return the distribution mode (which is the mean for Gaussian)."""
+        return self.mean
+    
+class VibeVoiceAcousticTokenizerModel(PreTrainedModel):
+    """VibeVoice speech tokenizer model combining encoder and decoder for acoustic tokens"""
+    
+    config_class = VibeVoiceAcousticTokenizerConfig
+    base_model_prefix = "vibevoice_acoustic_tokenizer"
+    _supports_flash_attn_2 = True  
+    _supports_sdpa = True  
+    _no_split_modules = ["TokenizerEncoder", "TokenizerDecoder"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        
+        self.register_buffer('fix_std', torch.tensor(config.fix_std), persistent=False)
+        self.std_dist_type = getattr(config, "std_dist_type", "fix")
+        
+        # Parse encoder depths
+        if isinstance(config.encoder_depths, str):
+            encoder_depths = [int(d) for d in config.encoder_depths.split('-')]
+        else:
+            encoder_depths = config.encoder_depths
+            
+        # Parse decoder depths if provided
+        if config.decoder_depths is not None and isinstance(config.decoder_depths, str):
+            decoder_depths = [int(d) for d in config.decoder_depths.split('-')]
+        else:
+            # Default: use reversed encoder depths if decoder_depths is None
+            decoder_depths = list(reversed(encoder_depths))
+        
+        # Create encoder config
+        encoder_config = copy.deepcopy(config)
+        encoder_config.dimension = config.vae_dim
+        encoder_config.n_filters = config.encoder_n_filters
+        encoder_config.ratios = config.encoder_ratios
+        encoder_config.depths = encoder_depths
+        encoder_config.norm = config.conv_norm
+        encoder_config.pad_mode = config.pad_mode
+        encoder_config.bias = config.conv_bias
+        encoder_config.layernorm_eps = config.layernorm_eps
+        encoder_config.layernorm_elementwise_affine = config.layernorm_elementwise_affine
+        encoder_config.mixer_layer = config.mixer_layer
+        encoder_config.layer_scale_init_value = config.layer_scale_init_value
+        encoder_config.disable_last_norm = config.disable_last_norm
+        
+        # Create decoder config
+        decoder_config = copy.deepcopy(config)
+        decoder_config.dimension = config.vae_dim
+        decoder_config.n_filters = config.decoder_n_filters
+        decoder_config.ratios = config.decoder_ratios
+        decoder_config.depths = decoder_depths
+        decoder_config.norm = config.conv_norm
+        decoder_config.pad_mode = config.pad_mode
+        decoder_config.bias = config.conv_bias
+        decoder_config.layernorm_eps = config.layernorm_eps
+        decoder_config.layernorm_elementwise_affine = config.layernorm_elementwise_affine
+        decoder_config.mixer_layer = config.mixer_layer
+        decoder_config.layer_scale_init_value = config.layer_scale_init_value
+        decoder_config.disable_last_norm = config.disable_last_norm
+        
+        # Initialize encoder and decoder
+        self.encoder = TokenizerEncoder(encoder_config)
+        self.decoder = TokenizerDecoder(decoder_config)
+        
+        # Initialize weights
+        self.apply(self._init_weights)
+    
+    def _init_weights(self, module):
+        """Initialize weights for the model"""
+        if isinstance(module, nn.Linear):
+            nn.init.normal_(module.weight, std=self.config.weight_init_value)
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+        elif isinstance(module, nn.LayerNorm):
+            nn.init.ones_(module.weight)
+            nn.init.zeros_(module.bias)
+        elif isinstance(module, nn.Conv1d):
+            nn.init.normal_(module.weight, std=self.config.weight_init_value)
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+    
+    @torch.no_grad()
+    def encode(self, audio, cache=None, sample_indices=None, use_cache=False, debug=False):
+        """Convert audio to latent representations"""
+        latents = self.encoder(audio, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        return VibeVoiceTokenizerEncoderOutput(mean=latents.permute(0, 2, 1), std=self.fix_std)
+    
+    @torch.no_grad()
+    def sampling(self, encoder_output, dist_type=None):
+        """Sample from the encoder output distribution"""
+        dist_type = dist_type or self.std_dist_type
+    
+        if dist_type == 'fix':
+            return encoder_output.sample(dist_type='fix')
+        elif dist_type == 'gaussian':
+            return encoder_output.sample(dist_type='gaussian')
+        else:
+            raise ValueError(f"Unsupported dist_type: {dist_type}, expected 'fix' or 'gaussian'")
+    
+    @torch.no_grad()
+    def decode(self, latents, cache=None, sample_indices=None, use_cache=False, debug=False):
+        """Convert latent representations back to audio"""
+        if latents.shape[1] == self.config.vae_dim:
+            pass
+        else:
+            latents = latents.permute(0, 2, 1)
+
+        audio = self.decoder(latents, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        return audio
+
+    def forward(self, audio, cache=None, sample_indices=None, use_cache=False, debug=False):
+        """Full forward pass: encode audio to latents, then decode back to audio"""
+        encoder_output = self.encode(audio, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        sampled_latents, _ = self.sampling(encoder_output)
+        reconstructed = self.decode(sampled_latents, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        return reconstructed, sampled_latents
+
+
+class VibeVoiceSemanticTokenizerModel(PreTrainedModel):
+    """VibeVoice speech tokenizer model with only encoder for semantic tokens"""
+    
+    config_class = VibeVoiceSemanticTokenizerConfig
+    base_model_prefix = "vibevoice_semantic_tokenizer"
+    _supports_flash_attn_2 = True  
+    _supports_sdpa = True  
+    _no_split_modules = ["TokenizerEncoder"]
+    
+    def __init__(self, config):
+        super().__init__(config)
+        
+        # Parse encoder depths
+        if isinstance(config.encoder_depths, str):
+            encoder_depths = [int(d) for d in config.encoder_depths.split('-')]
+        else:
+            encoder_depths = config.encoder_depths
+        
+        # Create encoder config
+        encoder_config = copy.deepcopy(config)
+        encoder_config.dimension = config.vae_dim
+        encoder_config.n_filters = config.encoder_n_filters
+        encoder_config.ratios = config.encoder_ratios
+        encoder_config.depths = encoder_depths
+        encoder_config.norm = config.conv_norm
+        encoder_config.pad_mode = config.pad_mode
+        encoder_config.bias = config.conv_bias
+        encoder_config.layernorm_eps = config.layernorm_eps
+        encoder_config.layernorm_elementwise_affine = config.layernorm_elementwise_affine
+        encoder_config.mixer_layer = config.mixer_layer
+        encoder_config.layer_scale_init_value = config.layer_scale_init_value
+        encoder_config.disable_last_norm = config.disable_last_norm
+        
+        # Initialize encoder and decoder
+        self.encoder = TokenizerEncoder(encoder_config)
+        
+        # Initialize weights
+        self.apply(self._init_weights)
+    
+    def _init_weights(self, module):
+        """Initialize weights for the model"""
+        if isinstance(module, nn.Linear):
+            nn.init.normal_(module.weight, std=self.config.weight_init_value)
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+        elif isinstance(module, nn.LayerNorm):
+            nn.init.ones_(module.weight)
+            nn.init.zeros_(module.bias)
+        elif isinstance(module, nn.Conv1d):
+            nn.init.normal_(module.weight, std=self.config.weight_init_value)
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+    
+    @torch.no_grad()
+    def encode(self, audio, cache=None, sample_indices=None, use_cache=False, debug=False):
+        """Convert audio to latent representations"""
+        latents = self.encoder(audio, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        return VibeVoiceTokenizerEncoderOutput(mean=latents.permute(0, 2, 1))
+    
+    @torch.no_grad()
+    def sampling(self, encoder_output, dist_type=None):
+        """Sample from the encoder output distribution"""
+        return encoder_output.sample(dist_type='none')
+
+    def forward(self, audio, cache=None, sample_indices=None, use_cache=False, debug=False):
+        """Full forward pass: encode audio to latents, then decode back to audio"""
+        encoder_output = self.encode(audio, cache=cache, sample_indices=sample_indices, use_cache=use_cache, debug=debug)
+        sampled_latents, _ = self.sampling(encoder_output, dist_type='none')
+        return None, sampled_latents
+
+AutoModel.register(VibeVoiceAcousticTokenizerConfig, VibeVoiceAcousticTokenizerModel)
+AutoModel.register(VibeVoiceSemanticTokenizerConfig, VibeVoiceSemanticTokenizerModel)
+
+__all__ = [
+    "VibeVoiceTokenizerStreamingCache",
+    "VibeVoiceAcousticTokenizerModel",
+    "VibeVoiceSemanticTokenizerModel",
+]

vvembed/modular/streamer.py

@@ -0,0 +1,264 @@
+from __future__ import annotations
+
+import torch
+
+import asyncio
+from queue import Queue
+from typing import TYPE_CHECKING, Optional
+
+
+from transformers.generation import BaseStreamer
+
+
+class AudioStreamer(BaseStreamer):
+    """
+    Audio streamer that stores audio chunks in queues for each sample in the batch.
+    This allows streaming audio generation for multiple samples simultaneously.
+    
+    Parameters:
+        batch_size (`int`):
+            The batch size for generation
+        stop_signal (`any`, *optional*):
+            The signal to put in the queue when generation ends. Defaults to None.
+        timeout (`float`, *optional*):
+            The timeout for the audio queue. If `None`, the queue will block indefinitely.
+    """
+    
+    def __init__(
+        self, 
+        batch_size: int,
+        stop_signal: Optional[any] = None,
+        timeout: Optional[float] = None,
+    ):
+        self.batch_size = batch_size
+        self.stop_signal = stop_signal
+        self.timeout = timeout
+        
+        # Create a queue for each sample in the batch
+        self.audio_queues = [Queue() for _ in range(batch_size)]
+        self.finished_flags = [False for _ in range(batch_size)]
+        self.sample_indices_map = {}  # Maps from sample index to queue index
+        
+    def put(self, audio_chunks: torch.Tensor, sample_indices: torch.Tensor):
+        """
+        Receives audio chunks and puts them in the appropriate queues.
+        
+        Args:
+            audio_chunks: Tensor of shape (num_samples, ...) containing audio chunks
+            sample_indices: Tensor indicating which samples these chunks belong to
+        """
+        for i, sample_idx in enumerate(sample_indices):
+            idx = sample_idx.item()
+            if idx < self.batch_size and not self.finished_flags[idx]:
+                # Convert to numpy or keep as tensor based on preference
+                audio_chunk = audio_chunks[i].detach().cpu()
+                self.audio_queues[idx].put(audio_chunk, timeout=self.timeout)
+    
+    def end(self, sample_indices: Optional[torch.Tensor] = None):
+        """
+        Signals the end of generation for specified samples or all samples.
+        
+        Args:
+            sample_indices: Optional tensor of sample indices to end. If None, ends all.
+        """
+        if sample_indices is None:
+            # End all samples
+            for idx in range(self.batch_size):
+                if not self.finished_flags[idx]:
+                    self.audio_queues[idx].put(self.stop_signal, timeout=self.timeout)
+                    self.finished_flags[idx] = True
+        else:
+            # End specific samples
+            for sample_idx in sample_indices:
+                idx = sample_idx.item() if torch.is_tensor(sample_idx) else sample_idx
+                if idx < self.batch_size and not self.finished_flags[idx]:
+                    self.audio_queues[idx].put(self.stop_signal, timeout=self.timeout)
+                    self.finished_flags[idx] = True
+    
+    def __iter__(self):
+        """Returns an iterator over the batch of audio streams."""
+        return AudioBatchIterator(self)
+    
+    def get_stream(self, sample_idx: int):
+        """Get the audio stream for a specific sample."""
+        if sample_idx >= self.batch_size:
+            raise ValueError(f"Sample index {sample_idx} exceeds batch size {self.batch_size}")
+        return AudioSampleIterator(self, sample_idx)
+
+
+class AudioSampleIterator:
+    """Iterator for a single audio stream from the batch."""
+    
+    def __init__(self, streamer: AudioStreamer, sample_idx: int):
+        self.streamer = streamer
+        self.sample_idx = sample_idx
+        
+    def __iter__(self):
+        return self
+    
+    def __next__(self):
+        value = self.streamer.audio_queues[self.sample_idx].get(timeout=self.streamer.timeout)
+        if value == self.streamer.stop_signal:
+            raise StopIteration()
+        return value
+
+
+class AudioBatchIterator:
+    """Iterator that yields audio chunks for all samples in the batch."""
+    
+    def __init__(self, streamer: AudioStreamer):
+        self.streamer = streamer
+        self.active_samples = set(range(streamer.batch_size))
+        
+    def __iter__(self):
+        return self
+    
+    def __next__(self):
+        if not self.active_samples:
+            raise StopIteration()
+            
+        batch_chunks = {}
+        samples_to_remove = set()
+        
+        # Try to get chunks from all active samples
+        for idx in self.active_samples:
+            try:
+                value = self.streamer.audio_queues[idx].get(block=False)
+                if value == self.streamer.stop_signal:
+                    samples_to_remove.add(idx)
+                else:
+                    batch_chunks[idx] = value
+            except:
+                # Queue is empty for this sample, skip it this iteration
+                pass
+        
+        # Remove finished samples
+        self.active_samples -= samples_to_remove
+        
+        if batch_chunks:
+            return batch_chunks
+        elif self.active_samples:
+            # If no chunks were ready but we still have active samples, 
+            # wait a bit and try again
+            import time
+            time.sleep(0.01)
+            return self.__next__()
+        else:
+            raise StopIteration()
+
+
+class AsyncAudioStreamer(AudioStreamer):
+    """
+    Async version of AudioStreamer for use in async contexts.
+    """
+    
+    def __init__(
+        self, 
+        batch_size: int,
+        stop_signal: Optional[any] = None,
+        timeout: Optional[float] = None,
+    ):
+        super().__init__(batch_size, stop_signal, timeout)
+        # Replace regular queues with async queues
+        self.audio_queues = [asyncio.Queue() for _ in range(batch_size)]
+        self.loop = asyncio.get_running_loop()
+        
+    def put(self, audio_chunks: torch.Tensor, sample_indices: torch.Tensor):
+        """Put audio chunks in the appropriate async queues."""
+        for i, sample_idx in enumerate(sample_indices):
+            idx = sample_idx.item()
+            if idx < self.batch_size and not self.finished_flags[idx]:
+                audio_chunk = audio_chunks[i].detach().cpu()
+                self.loop.call_soon_threadsafe(
+                    self.audio_queues[idx].put_nowait, audio_chunk
+                )
+    
+    def end(self, sample_indices: Optional[torch.Tensor] = None):
+        """Signal the end of generation for specified samples."""
+        if sample_indices is None:
+            indices_to_end = range(self.batch_size)
+        else:
+            indices_to_end = [s.item() if torch.is_tensor(s) else s for s in sample_indices]
+            
+        for idx in indices_to_end:
+            if idx < self.batch_size and not self.finished_flags[idx]:
+                self.loop.call_soon_threadsafe(
+                    self.audio_queues[idx].put_nowait, self.stop_signal
+                )
+                self.finished_flags[idx] = True
+    
+    async def get_stream(self, sample_idx: int):
+        """Get async iterator for a specific sample's audio stream."""
+        if sample_idx >= self.batch_size:
+            raise ValueError(f"Sample index {sample_idx} exceeds batch size {self.batch_size}")
+            
+        while True:
+            value = await self.audio_queues[sample_idx].get()
+            if value == self.stop_signal:
+                break
+            yield value
+    
+    def __aiter__(self):
+        """Returns an async iterator over all audio streams."""
+        return AsyncAudioBatchIterator(self)
+
+
+class AsyncAudioBatchIterator:
+    """Async iterator for batch audio streaming."""
+    
+    def __init__(self, streamer: AsyncAudioStreamer):
+        self.streamer = streamer
+        self.active_samples = set(range(streamer.batch_size))
+        
+    def __aiter__(self):
+        return self
+        
+    async def __anext__(self):
+        if not self.active_samples:
+            raise StopAsyncIteration()
+            
+        batch_chunks = {}
+        samples_to_remove = set()
+        
+        # Create tasks for all active samples
+        tasks = {
+            idx: asyncio.create_task(self._get_chunk(idx)) 
+            for idx in self.active_samples
+        }
+        
+        # Wait for at least one chunk to be ready
+        done, pending = await asyncio.wait(
+            tasks.values(), 
+            return_when=asyncio.FIRST_COMPLETED,
+            timeout=self.streamer.timeout
+        )
+        
+        # Cancel pending tasks
+        for task in pending:
+            task.cancel()
+            
+        # Process completed tasks
+        for idx, task in tasks.items():
+            if task in done:
+                try:
+                    value = await task
+                    if value == self.streamer.stop_signal:
+                        samples_to_remove.add(idx)
+                    else:
+                        batch_chunks[idx] = value
+                except asyncio.CancelledError:
+                    pass
+                    
+        self.active_samples -= samples_to_remove
+        
+        if batch_chunks:
+            return batch_chunks
+        elif self.active_samples:
+            # Try again if we still have active samples
+            return await self.__anext__()
+        else:
+            raise StopAsyncIteration()
+    
+    async def _get_chunk(self, idx):
+        """Helper to get a chunk from a specific queue."""
+        return await self.streamer.audio_queues[idx].get()

vvembed/processor/vibevoice_processor.py

@@ -0,0 +1,677 @@
+import math
+import warnings
+from typing import List, Optional, Union, Dict, Any, Tuple
+import os
+import re
+
+import numpy as np
+import torch
+
+from transformers.tokenization_utils_base import BatchEncoding, PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
+from transformers.utils import TensorType, logging
+from .vibevoice_tokenizer_processor import AudioNormalizer
+
+logger = logging.get_logger(__name__)
+
+
+class VibeVoiceProcessor:
+    r"""
+    Constructs a VibeVoice processor which wraps a VibeVoice tokenizer and audio processor into a single processor.
+
+    [`VibeVoiceProcessor`] offers all the functionalities of [`VibeVoiceTokenizer`] and [`VibeVoiceTokenizerProcessor`]. 
+    See the [`~VibeVoiceProcessor.__call__`] and [`~VibeVoiceProcessor.decode`] for more information.
+
+    Args:
+        tokenizer (`VibeVoiceTextTokenizer` or `VibeVoiceTextTokenizerFast`):
+            The tokenizer for text processing.
+        audio_processor (`VibeVoiceTokenizerProcessor`):
+            The audio processor for speech processing.
+        speech_tok_compress_ratio (`int`, *optional*, defaults to 3200):
+            The compression ratio for speech tokenization.
+        db_normalize (`bool`, *optional*, defaults to True):
+            Whether to apply decibel normalization to audio inputs.
+    """
+
+    def __init__(self, tokenizer=None, audio_processor=None, speech_tok_compress_ratio=3200, db_normalize=True, **kwargs):
+        self.tokenizer = tokenizer
+        self.audio_processor = audio_processor
+        self.speech_tok_compress_ratio = speech_tok_compress_ratio
+        self.db_normalize = db_normalize
+        self.audio_normalizer = AudioNormalizer() if db_normalize else None
+        self.system_prompt = " Transform the text provided by various speakers into speech output, utilizing the distinct voice of each respective speaker.\n"
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path, **kwargs):
+        """
+        Instantiate a VibeVoiceProcessor from a pretrained VibeVoice processor.
+
+        Args:
+            pretrained_model_name_or_path (`str` or `os.PathLike`):
+                This can be either:
+                - a string, the *model id* of a pretrained model
+                - a path to a *directory* containing processor config
+
+        Returns:
+            [`VibeVoiceProcessor`]: The processor object instantiated from pretrained model.
+        """
+        import os
+        import json
+        from .vibevoice_tokenizer_processor import VibeVoiceTokenizerProcessor
+        from modular.modular_vibevoice_text_tokenizer import (
+            VibeVoiceTextTokenizer, 
+            VibeVoiceTextTokenizerFast
+        )
+        
+        # Load processor configuration
+        config_path = os.path.join(pretrained_model_name_or_path, "preprocessor_config.json")
+        if os.path.exists(config_path):
+            with open(config_path, 'r') as f:
+                config = json.load(f)
+        else:
+            logger.warning(f"No preprocessor_config.json found at {pretrained_model_name_or_path}, using defaults")
+            config = {
+                "speech_tok_compress_ratio": 3200,
+                "db_normalize": True,
+            }
+        
+        # Extract main processor parameters
+        speech_tok_compress_ratio = config.get("speech_tok_compress_ratio", 3200)
+        db_normalize = config.get("db_normalize", True)
+        
+        # Load tokenizer - try from model path first, then fallback to Qwen        
+        language_model_pretrained_name = config.get("language_model_pretrained_name", None) or kwargs.pop("language_model_pretrained_name", "Qwen/Qwen2.5-1.5B")
+        logger.info(f"Loading tokenizer from {language_model_pretrained_name}")
+        if 'qwen' in language_model_pretrained_name.lower():
+            tokenizer = VibeVoiceTextTokenizerFast.from_pretrained(
+                language_model_pretrained_name,
+                **kwargs
+            )
+        else:
+            raise ValueError(f"Unsupported tokenizer type for {language_model_pretrained_name}. Supported types: Qwen, Llama, Gemma.")
+        
+        # Load audio processor
+        if "audio_processor" in config:
+            # Create audio processor from config
+            audio_config = config["audio_processor"]
+            audio_processor = VibeVoiceTokenizerProcessor(
+                sampling_rate=audio_config.get("sampling_rate", 24000),
+                normalize_audio=audio_config.get("normalize_audio", True),
+                target_dB_FS=audio_config.get("target_dB_FS", -25),
+                eps=audio_config.get("eps", 1e-6),
+            )
+        else:
+            # Create default audio processor
+            audio_processor = VibeVoiceTokenizerProcessor()
+        
+        # Create and return the processor
+        return cls(
+            tokenizer=tokenizer,
+            audio_processor=audio_processor,
+            speech_tok_compress_ratio=speech_tok_compress_ratio,
+            db_normalize=db_normalize,
+        )
+    
+    def save_pretrained(self, save_directory: Union[str, os.PathLike], **kwargs):
+        """
+        Save a processor to a directory, so that it can be re-loaded using the
+        [`~VibeVoiceProcessor.from_pretrained`] class method.
+
+        Args:
+            save_directory (`str` or `os.PathLike`):
+                Directory where the processor will be saved.
+        """
+        import os
+        import json
+        
+        os.makedirs(save_directory, exist_ok=True)
+        
+        # Save processor configuration
+        processor_config = {
+            "processor_class": "VibeVoiceProcessor",
+            "speech_tok_compress_ratio": self.speech_tok_compress_ratio,
+            "db_normalize": self.db_normalize,
+            "audio_processor": {
+                "feature_extractor_type": "VibeVoiceTokenizerProcessor",
+                "sampling_rate": getattr(self.audio_processor, 'sampling_rate', 24000),
+                "normalize_audio": getattr(self.audio_processor, 'normalize_audio', True),
+                "target_dB_FS": getattr(self.audio_processor, 'target_dB_FS', -25),
+                "eps": getattr(self.audio_processor, 'eps', 1e-6),
+            }
+        }
+        
+        config_path = os.path.join(save_directory, "preprocessor_config.json")
+        with open(config_path, 'w') as f:
+            json.dump(processor_config, f, indent=2)
+        
+        logger.info(f"Processor configuration saved in {config_path}")
+    
+    def __call__(
+        self,
+        text: Optional[Union[str, List[str], TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]]] = None,
+        voice_samples: Optional[Union[List[Union[str, np.ndarray]], List[List[Union[str, np.ndarray]]]]] = None,
+        padding: Union[bool, str, PaddingStrategy] = True,
+        truncation: Union[bool, str, TruncationStrategy] = False,
+        max_length: Optional[int] = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        return_attention_mask: bool = True,
+        **kwargs,
+    ) -> BatchEncoding:
+        """
+        Main method to process one or more podcast scripts with optional voice samples.
+
+        Args:
+            text (`str`, `List[str]`):
+                The input text(s) to process. Can be:
+                - A single script string
+                - A list of script strings for batch processing
+                - A path to a .json or .txt file
+                - A list of paths
+            voice_samples (`List[Union[str, np.ndarray]]`, `List[List[Union[str, np.ndarray]]]`, *optional*):
+                Voice samples for each script. Can be:
+                - A list of samples for a single script
+                - A list of lists for batch processing
+            padding (`bool`, `str` or `PaddingStrategy`, defaults to `True`):
+                Whether to pad sequences to the same length
+            truncation (`bool`, `str` or `TruncationStrategy`, defaults to `False`):
+                Whether to truncate sequences
+            max_length (`int`, *optional*):
+                Maximum length of the returned sequences
+            return_tensors (`str` or `TensorType`, *optional*):
+                If set, will return tensors of a particular framework
+            return_attention_mask (`bool`, defaults to `True`):
+                Whether to return the attention mask
+
+        Returns:
+            `BatchEncoding`: A BatchEncoding with the following fields:
+                - **input_ids** -- List of token id sequences or tensor
+                - **attention_mask** -- List of attention masks or tensor
+                - **speech_tensors** -- Padded speech inputs (if voice_samples provided)
+                - **speech_masks** -- Speech masks (if voice_samples provided)
+                - **speech_input_mask** -- Boolean masks indicating speech token positions
+        """
+        # Handle single vs batch input
+        if isinstance(text, str) or (isinstance(text, list) and len(text) > 0 and not isinstance(text[0], str)):
+            # Single input
+            texts = [text]
+            is_batched = False
+        else:
+            # Batch input
+            texts = text
+            is_batched = True
+            
+        # Handle voice samples
+        if voice_samples is not None:
+            if not is_batched or (isinstance(voice_samples[0], (str, np.ndarray))):
+                # Single set of voice samples
+                voice_samples_list = [voice_samples]
+            else:
+                # Batch of voice samples
+                voice_samples_list = voice_samples
+        else:
+            voice_samples_list = [None] * len(texts)
+        
+        # Process each input
+        all_encodings = []
+        for text_input, voice_input in zip(texts, voice_samples_list):
+            encoding = self._process_single(text_input, voice_input)
+            all_encodings.append(encoding)
+            
+        # Combine batch
+        batch_encoding = self._batch_encode(
+            all_encodings,
+            padding=padding,
+            truncation=truncation,
+            max_length=max_length,
+            return_tensors=return_tensors,
+            return_attention_mask=return_attention_mask,
+        )
+        
+        return batch_encoding
+    
+    def _process_single(
+        self,
+        text: Union[str, TextInput],
+        voice_samples: Optional[List[Union[str, np.ndarray]]] = None,
+    ) -> Dict[str, Any]:
+        """Process a single podcast script."""
+        # Determine if text is a file path or direct script
+        script = None
+        if isinstance(text, str):
+            # Check if it's a file path
+            if text.endswith('.json') and os.path.exists(text):
+                script = self._convert_json_to_script(text)
+            elif text.endswith('.txt') and os.path.exists(text):
+                script = self._convert_text_to_script(text)
+            else:
+                # Assume it's the script content directly
+                script = text
+        
+        if script is None:
+            raise ValueError(f"Could not process input text: {text}")
+        
+        # Parse the script
+        parsed_lines = self._parse_script(script)
+        all_speakers = list(set(speaker_id for speaker_id, _ in parsed_lines))
+        
+        # Create system prompt
+        # system_tokens = self.tokenizer.encode(self.system_prompt, add_special_tokens=False)
+        system_tokens = self.tokenizer.encode(self.system_prompt)
+        
+        # Process voice samples if provided
+        if voice_samples:
+            voice_tokens, voice_speech_inputs, voice_speech_masks = self._create_voice_prompt(voice_samples[:len(all_speakers)])
+        else:
+            voice_tokens, voice_speech_inputs, voice_speech_masks = [], [], []
+        
+        # Build full token sequence
+        full_tokens = system_tokens + voice_tokens
+        speech_input_mask = [False] * len(system_tokens) + voice_speech_masks
+        
+        # Add text input section
+        full_tokens += self.tokenizer.encode(' Text input:\n', add_special_tokens=False)
+        speech_input_mask += [False] * len(self.tokenizer.encode(' Text input:\n', add_special_tokens=False))
+        
+        for speaker_id, speaker_text in parsed_lines:
+            speaker_text_tokens = self.tokenizer.encode(f" Speaker {speaker_id}:{speaker_text}\n", add_special_tokens=False)
+            full_tokens += speaker_text_tokens
+            speech_input_mask += [False] * len(speaker_text_tokens)
+        
+        # Add speech output section
+        full_tokens += self.tokenizer.encode(' Speech output:\n', add_special_tokens=False) + [self.tokenizer.speech_start_id]
+        speech_input_mask += [False] * (len(self.tokenizer.encode(' Speech output:\n', add_special_tokens=False)) + 1)
+        
+        return {
+            "input_ids": full_tokens,
+            "speech_inputs": voice_speech_inputs if voice_speech_inputs else None,
+            "speech_input_mask": speech_input_mask,
+            "parsed_script": parsed_lines,
+            "all_speakers": all_speakers,
+        }
+    
+    def _batch_encode(
+        self,
+        encodings: List[Dict[str, Any]],
+        padding: Union[bool, str, PaddingStrategy] = True,
+        truncation: Union[bool, str, TruncationStrategy] = False,
+        max_length: Optional[int] = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        return_attention_mask: bool = True,
+    ) -> BatchEncoding:
+        """Combine multiple encodings into a batch with padding."""
+        # Extract input_ids and create attention_mask
+        input_ids_list = [enc["input_ids"] for enc in encodings]
+        speech_input_masks_list = [enc["speech_input_mask"] for enc in encodings]
+        
+        # Determine padding strategy
+        if isinstance(padding, bool):
+            padding_strategy = PaddingStrategy.LONGEST if padding else PaddingStrategy.DO_NOT_PAD
+        elif isinstance(padding, str):
+            padding_strategy = PaddingStrategy(padding)
+        else:
+            padding_strategy = padding
+            
+        # Apply padding to input_ids
+        if padding_strategy != PaddingStrategy.DO_NOT_PAD:
+            if padding_strategy == PaddingStrategy.LONGEST:
+                max_len = max(len(ids) for ids in input_ids_list)
+            elif padding_strategy == PaddingStrategy.MAX_LENGTH and max_length is not None:
+                max_len = max_length
+            else:
+                max_len = max(len(ids) for ids in input_ids_list)
+                
+            # Pad sequences
+            padded_input_ids = []
+            attention_masks = []
+            padded_speech_input_masks = []
+            
+            for input_ids, speech_mask in zip(input_ids_list, speech_input_masks_list):
+                # Truncate if needed
+                if truncation and len(input_ids) > max_len:
+                    input_ids = input_ids[:max_len]
+                    speech_mask = speech_mask[:max_len]
+                    
+                # Pad
+                padding_length = max_len - len(input_ids)
+                # padded_ids = [self.tokenizer.pad_token_id] * padding_length + input_ids
+                padded_ids = [self.tokenizer.pad_id] * padding_length + input_ids
+                attention_mask = [0] * padding_length + [1] * len(input_ids)
+                padded_speech_mask = [False] * padding_length + speech_mask
+                
+                padded_input_ids.append(padded_ids)
+                attention_masks.append(attention_mask)
+                padded_speech_input_masks.append(padded_speech_mask)
+                
+            input_ids_list = padded_input_ids
+            speech_input_masks_list = padded_speech_input_masks
+        else:
+            # No padding, just create attention masks
+            attention_masks = [[1] * len(ids) for ids in input_ids_list] if return_attention_mask else None
+            
+        # Process speech inputs
+        all_speech_inputs = []
+        has_speech = False
+        for enc in encodings:
+            if enc["speech_inputs"] is not None:
+                all_speech_inputs.extend(enc["speech_inputs"])
+                has_speech = True
+                
+        # Prepare batch encoding
+        batch_encoding = BatchEncoding()
+        
+        # Handle tensor conversion
+        if return_tensors is not None:
+            batch_encoding["input_ids"] = torch.tensor(input_ids_list, dtype=torch.long)
+            if return_attention_mask and attention_masks is not None:
+                batch_encoding["attention_mask"] = torch.tensor(attention_masks, dtype=torch.long)
+            batch_encoding["speech_input_mask"] = torch.tensor(speech_input_masks_list, dtype=torch.bool)
+        else:
+            batch_encoding["input_ids"] = input_ids_list
+            if return_attention_mask and attention_masks is not None:
+                batch_encoding["attention_mask"] = attention_masks
+            batch_encoding["speech_input_mask"] = speech_input_masks_list
+            
+        # Process speech tensors if present
+        if has_speech:
+            speech_dict = self.prepare_speech_inputs(
+                all_speech_inputs,
+                return_tensors=return_tensors,
+            )
+            batch_encoding["speech_tensors"] = speech_dict["padded_speeches"]
+            batch_encoding["speech_masks"] = speech_dict["speech_masks"]
+        else:
+            batch_encoding["speech_tensors"] = None
+            batch_encoding["speech_masks"] = None
+            
+        # Add metadata
+        batch_encoding["parsed_scripts"] = [enc["parsed_script"] for enc in encodings]
+        batch_encoding["all_speakers_list"] = [enc["all_speakers"] for enc in encodings]
+        
+        return batch_encoding
+
+    def _create_voice_prompt(
+        self, 
+        speaker_samples: List[Union[str, np.ndarray]]
+    ) -> Tuple[List[int], List[np.ndarray], List[bool]]:
+        """
+        Create voice prompt tokens and process audio samples.
+        
+        Returns:
+            tuple: (voice_tokens, voice_speech_inputs, voice_speech_masks)
+        """
+        vae_token_id = self.tokenizer.speech_diffusion_id
+        
+        voice_full_tokens = self.tokenizer.encode(' Voice input:\n', add_special_tokens=False)
+        voice_speech_inputs = []
+        voice_speech_masks = [False] * len(voice_full_tokens)
+        
+        for speaker_id, speaker_audio in enumerate(speaker_samples):
+            prefix_tokens = self.tokenizer.encode(f" Speaker {speaker_id}:", add_special_tokens=False)
+            
+            # Process audio
+            if isinstance(speaker_audio, str):
+                # Load audio from file
+                wav = self.audio_processor._load_audio_from_path(speaker_audio)
+            else:
+                wav = np.array(speaker_audio, dtype=np.float32)
+            
+            # Apply normalization if needed
+            if self.db_normalize and self.audio_normalizer:
+                wav = self.audio_normalizer(wav)
+            
+            # Calculate token length based on compression ratio
+            # if speaker_audio.endswith('.pt') or speaker_audio.endswith('.npy'):
+            #     vae_tok_len = wav.shape[0]
+            # else:
+            vae_tok_len = math.ceil(wav.shape[0] / self.speech_tok_compress_ratio)
+            
+            # Build tokens and masks
+            speaker_tokens = (prefix_tokens + 
+                            [self.tokenizer.speech_start_id] + 
+                            [vae_token_id] * vae_tok_len + 
+                            [self.tokenizer.speech_end_id] + 
+                            self.tokenizer.encode('\n', add_special_tokens=False))
+            
+            vae_input_mask = ([False] * len(prefix_tokens) + 
+                            [False] + 
+                            [True] * vae_tok_len + 
+                            [False] + 
+                            [False])
+            
+            voice_full_tokens.extend(speaker_tokens)
+            voice_speech_masks.extend(vae_input_mask)
+            voice_speech_inputs.append(wav)
+            
+        return voice_full_tokens, voice_speech_inputs, voice_speech_masks
+
+    def prepare_speech_inputs(
+        self,
+        speech_inputs: List[np.ndarray],
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        device: Optional[Union[str, torch.device]] = None,
+        dtype: Optional[torch.dtype] = None,
+    ) -> Dict[str, Any]:
+        """
+        Prepare speech inputs for model consumption.
+        
+        Args:
+            speech_inputs: List of speech arrays
+            return_tensors: Output tensor type
+            device: Device to place tensors on
+            dtype: Data type for tensors
+            
+        Returns:
+            Dictionary with padded_speeches and speech_masks
+        """
+        if not speech_inputs:
+            return {"padded_speeches": None, "speech_masks": None}
+        
+        # Calculate sequence lengths
+        vae_tok_seqlens = [math.ceil(s.shape[0] / self.speech_tok_compress_ratio) for s in speech_inputs]
+        # vae_tok_seqlens = [math.ceil(s.shape[0] / self.speech_tok_compress_ratio) if s.ndim == 1 else s.shape[0] for s in speech_inputs]
+        max_speech_length = max(s.shape[0] for s in speech_inputs)
+        
+        # Pad speeches
+        if speech_inputs[0].ndim == 1:
+            padded_speeches = np.full((len(speech_inputs), max_speech_length), fill_value=0, dtype=np.float32)
+        else:
+            padded_speeches = np.full((len(speech_inputs), max_speech_length, speech_inputs[0].shape[-1]), fill_value=0, dtype=np.float32)
+        speech_masks = np.zeros((len(speech_inputs), max(vae_tok_seqlens)), dtype=np.bool_)
+        
+        for i, (speech, vae_tok_length) in enumerate(zip(speech_inputs, vae_tok_seqlens)):
+            padded_speeches[i, :len(speech)] = speech
+            speech_masks[i, :vae_tok_length] = True
+        
+        result = {
+            "padded_speeches": padded_speeches,
+            "speech_masks": speech_masks,
+        }
+        
+        # Convert to tensors if requested
+        if return_tensors == "pt":
+            result["padded_speeches"] = torch.tensor(padded_speeches, device=device, dtype=dtype or torch.float32)
+            result["speech_masks"] = torch.tensor(speech_masks, device=device, dtype=torch.bool)
+            
+        return result
+        
+    def _convert_json_to_script(self, json_file: str) -> str:
+        """
+        Convert JSON format to script format.
+        Expected JSON format:
+        [
+            {"speaker": "1", "text": "Hello everyone..."},
+            {"speaker": "2", "text": "Great to be here..."}
+        ]
+        """
+        import json
+        
+        with open(json_file, 'r', encoding='utf-8') as f:
+            data = json.load(f)
+        
+        if not isinstance(data, list):
+            raise ValueError("JSON file must contain a list of speaker entries")
+        
+        script_lines = []
+        for item in data:
+            if not isinstance(item, dict):
+                logger.warning(f"Skipping non-dict entry: {item}")
+                continue
+                
+            speaker = item.get('speaker')
+            text = item.get('text')
+            
+            if speaker is None or text is None:
+                logger.warning(f"Skipping entry missing speaker or text: {item}")
+                continue
+            
+            # Ensure speaker ID is valid
+            try:
+                speaker_id = int(speaker)
+            except (ValueError, TypeError):
+                logger.warning(f"Invalid speaker ID: {speaker}, skipping entry")
+                continue
+            
+            # Clean up text
+            text = text.strip()
+            if text:
+                script_lines.append(f"Speaker {speaker_id}: {text}")
+        
+        if not script_lines:
+            raise ValueError("No valid entries found in JSON file")
+            
+        return "\n".join(script_lines)
+
+    def _convert_text_to_script(self, text_file: str) -> str:
+        """
+        Convert text file to script format.
+        Handles multiple formats:
+        1. Already formatted as "Speaker X: text"
+        2. Plain text (assigns to Speaker 1)
+        
+        Handles edge cases like multiple colons in a line.
+        """
+        with open(text_file, 'r', encoding='utf-8') as f:
+            lines = f.readlines()
+        
+        script_lines = []
+        current_speaker = 1
+        
+        for line in lines:
+            line = line.strip()
+            if not line:
+                continue
+            
+            # Try to parse as "Speaker X: text" format
+            # Use regex to be more robust
+            speaker_match = re.match(r'^Speaker\s+(\d+)\s*:\s*(.*)$', line, re.IGNORECASE)
+            
+            if speaker_match:
+                speaker_id = int(speaker_match.group(1))
+                text = speaker_match.group(2).strip()
+                if text:
+                    script_lines.append(f"Speaker {speaker_id}: {text}")
+            else:
+                # Treat as plain text - assign to current speaker
+                script_lines.append(f"Speaker {current_speaker}: {line}")
+        
+        if not script_lines:
+            raise ValueError("No valid content found in text file")
+            
+        return "\n".join(script_lines)
+
+    def _parse_script(self, script: str) -> List[Tuple[int, str]]:
+        """Parse script into list of (speaker_id, text) tuples."""
+        lines = script.strip().split("\n")
+        parsed_lines = []
+        speaker_ids = []
+                
+        # First pass: parse all lines and collect speaker IDs
+        for line in lines:
+            if not line.strip():
+                continue
+                
+            # Use regex to handle edge cases like multiple colons
+            match = re.match(r'^Speaker\s+(\d+)\s*:\s*(.*)$', line.strip(), re.IGNORECASE)
+            
+            if match:
+                speaker_id = int(match.group(1))
+                text = ' ' + match.group(2).strip()
+                parsed_lines.append((speaker_id, text))
+                speaker_ids.append(speaker_id)
+            else:
+                logger.warning(f"Could not parse line: '{line}'")
+        
+        if not parsed_lines:
+            raise ValueError("No valid speaker lines found in script")
+        
+        # Check if we need to normalize speaker IDs (only if all are > 0)
+        min_speaker_id = min(speaker_ids)
+        if min_speaker_id > 0:
+            # Normalize to start from 0
+            normalized_lines = []
+            for speaker_id, text in parsed_lines:
+                normalized_lines.append((speaker_id - 1, text))
+            return normalized_lines
+        else:
+            # Keep original IDs
+            return parsed_lines
+
+    def _merge_inputs(self, text_inputs: BatchEncoding, audio_inputs: Dict) -> BatchEncoding:
+        """Merge text and audio inputs into a single BatchEncoding."""
+        # Start with text inputs
+        merged = BatchEncoding(text_inputs)
+        
+        # Add audio-specific fields
+        if "audio" in audio_inputs:
+            merged["speech_inputs"] = audio_inputs["audio"]
+        if "streaming" in audio_inputs:
+            merged["streaming"] = audio_inputs["streaming"]
+            
+        return merged
+
+    def batch_decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to VibeVoiceTextTokenizer's [`~PreTrainedTokenizer.batch_decode`].
+        Please refer to the docstring of this method for more information.
+        """
+        return self.tokenizer.batch_decode(*args, **kwargs)
+
+    def decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to VibeVoiceTextTokenizer's [`~PreTrainedTokenizer.decode`].
+        Please refer to the docstring of this method for more information.
+        """
+        return self.tokenizer.decode(*args, **kwargs)
+
+    @property
+    def model_input_names(self):
+        """
+        Return the list of inputs accepted by the model.
+        """
+        tokenizer_input_names = self.tokenizer.model_input_names
+        audio_processor_input_names = self.audio_processor.model_input_names
+        return list(dict.fromkeys(tokenizer_input_names + audio_processor_input_names + ["speech_inputs", "speech_input_mask"]))
+
+    def save_audio(self, 
+        audio: Union[torch.Tensor, np.ndarray, List[Union[torch.Tensor, np.ndarray]]],
+        output_path: str = "output.wav",
+        sampling_rate: Optional[int] = None,
+        normalize: bool = False,
+        batch_prefix: str = "audio_",
+    ) -> str:
+        """
+        Save audio data to a file.
+        Args:
+            audio (Union[torch.Tensor, np.ndarray, List[Union[torch.Tensor, np.ndarray]]]):
+                The audio data to save. Can be a single tensor/array or a list of them.
+            output_path (str, optional): Path to save the audio file. Defaults to "output.wav".
+            sampling_rate (int, optional): Sampling rate for the audio. If None, uses the processor's default.
+            normalize (bool, optional): Whether to normalize the audio before saving. Defaults to False.
+            batch_prefix (str, optional): Prefix for batch audio files. Defaults to "audio_".
+        Returns:
+            str: The path to the saved audio file.
+        """
+        return self.audio_processor.save_audio(audio, output_path=output_path, sampling_rate=sampling_rate, normalize=normalize, batch_prefix=batch_prefix)
+    
+__all__ = [
+    "VibeVoiceProcessor",
+]

vvembed/processor/vibevoice_tokenizer_processor.py

@@ -0,0 +1,483 @@
+"""
+Processor class for VibeVoice models.
+"""
+
+import os
+import json
+import warnings
+from typing import List, Optional, Union, Dict, Any
+
+import numpy as np
+import torch
+
+from transformers.feature_extraction_utils import FeatureExtractionMixin
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+
+class AudioNormalizer:
+    """
+    Audio normalization class for VibeVoice tokenizer.
+    
+    This class provides audio normalization to ensure consistent input levels
+    for the VibeVoice tokenizer while maintaining audio quality.
+    """
+    
+    def __init__(self, target_dB_FS: float = -25, eps: float = 1e-6):
+        """
+        Initialize the audio normalizer.
+        
+        Args:
+            target_dB_FS (float): Target dB FS level for the audio. Default: -25
+            eps (float): Small value to avoid division by zero. Default: 1e-6
+        """
+        self.target_dB_FS = target_dB_FS
+        self.eps = eps
+    
+    def tailor_dB_FS(self, audio: np.ndarray) -> tuple:
+        """
+        Adjust the audio to the target dB FS level.
+        
+        Args:
+            audio (np.ndarray): Input audio signal
+            
+        Returns:
+            tuple: (normalized_audio, rms, scalar)
+        """
+        rms = np.sqrt(np.mean(audio**2))
+        scalar = 10 ** (self.target_dB_FS / 20) / (rms + self.eps)
+        normalized_audio = audio * scalar
+        return normalized_audio, rms, scalar
+    
+    def avoid_clipping(self, audio: np.ndarray, scalar: Optional[float] = None) -> tuple:
+        """
+        Avoid clipping by scaling down if necessary.
+        
+        Args:
+            audio (np.ndarray): Input audio signal
+            scalar (float, optional): Explicit scaling factor
+            
+        Returns:
+            tuple: (normalized_audio, scalar)
+        """
+        if scalar is None:
+            max_val = np.max(np.abs(audio))
+            if max_val > 1.0:
+                scalar = max_val + self.eps
+            else:
+                scalar = 1.0
+        
+        return audio / scalar, scalar
+    
+    def __call__(self, audio: np.ndarray) -> np.ndarray:
+        """
+        Normalize the audio by adjusting to target dB FS and avoiding clipping.
+        
+        Args:
+            audio (np.ndarray): Input audio signal
+            
+        Returns:
+            np.ndarray: Normalized audio signal
+        """
+        # First adjust to target dB FS
+        audio, _, _ = self.tailor_dB_FS(audio)
+        # Then avoid clipping
+        audio, _ = self.avoid_clipping(audio)
+        return audio
+
+
+# Change from ProcessorMixin to FeatureExtractionMixin which is designed for single components
+class VibeVoiceTokenizerProcessor(FeatureExtractionMixin):
+    """
+    Processor for VibeVoice acoustic tokenizer models.
+    
+    This processor handles audio preprocessing for VibeVoice models, including:
+    - Audio format conversion (stereo to mono)
+    - Optional audio normalization
+    - Streaming support for infinite-length audio
+    
+    Args:
+        sampling_rate (int, optional): Expected sampling rate. Defaults to 24000.
+        normalize_audio (bool, optional): Whether to normalize audio. Defaults to True.
+        target_dB_FS (float, optional): Target dB FS for normalization. Defaults to -25.
+        eps (float, optional): Small value for numerical stability. Defaults to 1e-6.
+    """
+    model_input_names = ["input_features"]
+    
+    def __init__(
+        self,
+        sampling_rate: int = 24000,
+        normalize_audio: bool = True,
+        target_dB_FS: float = -25,
+        eps: float = 1e-6,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        
+        self.sampling_rate = sampling_rate
+        self.normalize_audio = normalize_audio
+        
+        # Initialize audio normalizer if needed
+        if self.normalize_audio:
+            self.normalizer = AudioNormalizer(target_dB_FS=target_dB_FS, eps=eps)
+        else:
+            self.normalizer = None
+        
+        # Save config
+        self.feature_extractor_dict = {
+            "sampling_rate": sampling_rate,
+            "normalize_audio": normalize_audio,
+            "target_dB_FS": target_dB_FS,
+            "eps": eps,
+        }
+    
+    def _ensure_mono(self, audio: np.ndarray) -> np.ndarray:
+        """
+        Convert stereo audio to mono if needed.
+        
+        Args:
+            audio (np.ndarray): Input audio array
+            
+        Returns:
+            np.ndarray: Mono audio array
+        """
+        if len(audio.shape) == 1:
+            return audio
+        elif len(audio.shape) == 2:
+            if audio.shape[0] == 2:  # (2, time)
+                return np.mean(audio, axis=0)
+            elif audio.shape[1] == 2:  # (time, 2)
+                return np.mean(audio, axis=1)
+            else:
+                # If one dimension is 1, squeeze it
+                if audio.shape[0] == 1:
+                    return audio.squeeze(0)
+                elif audio.shape[1] == 1:
+                    return audio.squeeze(1)
+                else:
+                    raise ValueError(f"Unexpected audio shape: {audio.shape}")
+        else:
+            raise ValueError(f"Audio should be 1D or 2D, got shape: {audio.shape}")
+    
+    def _process_single_audio(self, audio: Union[np.ndarray, List[float]]) -> np.ndarray:
+        """
+        Process a single audio array.
+        
+        Args:
+            audio: Single audio input
+            
+        Returns:
+            np.ndarray: Processed audio
+        """
+        # Convert to numpy array
+        if not isinstance(audio, np.ndarray):
+            audio = np.array(audio, dtype=np.float32)
+        else:
+            audio = audio.astype(np.float32)
+        
+        # Ensure mono
+        audio = self._ensure_mono(audio)
+        
+        # Normalize if requested
+        if self.normalize_audio and self.normalizer is not None:
+            audio = self.normalizer(audio)
+        
+        return audio
+    
+    def __call__(
+        self,
+        audio: Union[str, np.ndarray, List[float], List[np.ndarray], List[List[float]], List[str]] = None,
+        sampling_rate: Optional[int] = None,
+        return_tensors: Optional[str] = None,
+        **kwargs,
+    ):
+        """
+        Process audio for VibeVoice models.
+        
+        Args:
+            audio: Audio input(s) to process. Can be:
+                - str: Path to audio file
+                - np.ndarray: Audio array
+                - List[float]: Audio as list of floats
+                - List[np.ndarray]: Batch of audio arrays
+                - List[str]: Batch of audio file paths
+            sampling_rate (int, optional): Sampling rate of the input audio
+            return_tensors (str, optional): Return format ('pt' for PyTorch, 'np' for NumPy)
+            
+        Returns:
+            dict: Processed audio inputs with keys:
+                - input_features: Audio tensor(s) ready for the model
+        """
+        if audio is None:
+            raise ValueError("Audio input is required")
+        
+        # Validate sampling rate
+        if sampling_rate is not None and sampling_rate != self.sampling_rate:
+            logger.warning(
+                f"Input sampling rate ({sampling_rate}) differs from expected "
+                f"sampling rate ({self.sampling_rate}). Please resample your audio."
+            )
+        
+        # Handle different input types
+        if isinstance(audio, str):
+            # Single audio file path
+            audio = self._load_audio_from_path(audio)
+            is_batched = False
+        elif isinstance(audio, list):
+            if len(audio) == 0:
+                raise ValueError("Empty audio list provided")
+            
+            # Check if it's a list of file paths
+            if all(isinstance(item, str) for item in audio):
+                # Batch of audio file paths
+                audio = [self._load_audio_from_path(path) for path in audio]
+                is_batched = True
+            else:
+                # Check if it's batched audio arrays
+                is_batched = isinstance(audio[0], (np.ndarray, list))
+        else:
+            # Single audio array or list
+            is_batched = False
+        
+        # Process audio
+        if is_batched:
+            processed_audio = [self._process_single_audio(a) for a in audio]
+        else:
+            processed_audio = [self._process_single_audio(audio)]
+        
+        # Convert to tensors if requested
+        if return_tensors == "pt":
+            if len(processed_audio) == 1:
+                # Create a proper batch dimension (B, T)
+                input_features = torch.from_numpy(processed_audio[0]).unsqueeze(0).unsqueeze(1)
+            else:
+                # For batched input with different lengths, create a batch properly
+                input_features = torch.stack([torch.from_numpy(a) for a in processed_audio]).unsqueeze(1)
+        elif return_tensors == "np":
+            if len(processed_audio) == 1:
+                input_features = processed_audio[0][np.newaxis, np.newaxis, :]
+            else:
+                input_features = np.stack(processed_audio)[:, np.newaxis, :]
+        else:
+            input_features = processed_audio[0] if len(processed_audio) == 1 else processed_audio
+        
+        outputs = {
+            "audio": input_features,  # Use "audio" instead of "input_features"
+        }
+        
+        return outputs
+
+    def _load_audio_from_path(self, audio_path: str) -> np.ndarray:
+        """
+        Load audio from file path.
+        
+        Args:
+            audio_path (str): Path to audio file
+            
+        Returns:
+            np.ndarray: Loaded audio array
+        """
+        # Get file extension to determine loading method
+        file_ext = os.path.splitext(audio_path)[1].lower()
+        
+        if file_ext in ['.wav', '.mp3', '.flac', '.m4a', '.ogg']:
+            # Audio file - use librosa
+            import librosa
+            audio_array, sr = librosa.load(
+                audio_path, 
+                sr=self.sampling_rate, 
+                mono=True
+            )
+            return audio_array
+        elif file_ext == '.pt':
+            # PyTorch tensor file
+            audio_tensor = torch.load(audio_path, map_location='cpu').squeeze()
+            if isinstance(audio_tensor, torch.Tensor):
+                audio_array = audio_tensor.numpy()
+            else:
+                audio_array = np.array(audio_tensor)
+            return audio_array.astype(np.float32)
+        elif file_ext == '.npy':
+            # NumPy file
+            audio_array = np.load(audio_path)
+            return audio_array.astype(np.float32)
+        else:
+            raise ValueError(
+                f"Unsupported file format: {file_ext}. "
+                f"Supported formats: .wav, .mp3, .flac, .m4a, .ogg, .pt, .npy, .npz"
+            )
+    
+    def preprocess_audio(
+        self, 
+        audio_path_or_array: Union[str, np.ndarray],
+        normalize: Optional[bool] = None,
+    ) -> np.ndarray:
+        """
+        Convenience method to preprocess audio from file path or array.
+        This method is kept for backward compatibility but __call__ is recommended.
+        
+        Args:
+            audio_path_or_array: Path to audio file or numpy array
+            normalize: Whether to normalize (overrides default setting)
+            
+        Returns:
+            np.ndarray: Preprocessed audio array
+        """
+        if isinstance(audio_path_or_array, str):
+            audio_array = self._load_audio_from_path(audio_path_or_array)
+        else:
+            audio_array = np.array(audio_path_or_array, dtype=np.float32)
+        
+        # Override normalization setting if specified
+        original_normalize = self.normalize_audio
+        if normalize is not None:
+            self.normalize_audio = normalize
+        
+        try:
+            processed = self._process_single_audio(audio_array)
+        finally:
+            # Restore original setting
+            self.normalize_audio = original_normalize
+        
+        return processed
+    
+    # Override to_dict method for configuration saving
+    def to_dict(self) -> Dict[str, Any]:
+        """
+        Convert the object to a dict containing all attributes needed for serialization.
+        """
+        return self.feature_extractor_dict
+
+    def save_audio(
+        self,
+        audio: Union[torch.Tensor, np.ndarray, List[Union[torch.Tensor, np.ndarray]]],
+        output_path: str = "output.wav",
+        sampling_rate: Optional[int] = None,
+        normalize: bool = False,
+        batch_prefix: str = "audio_",
+    ):
+        """
+        Save audio data to WAV file(s).
+        
+        Args:
+            audio: Audio data to save. Can be:
+                - torch.Tensor: PyTorch tensor with shape (B, C, T) or (B, T) or (T)
+                - np.ndarray: NumPy array with shape (B, C, T) or (B, T) or (T)
+                - List of tensors or arrays
+            output_path: Path where to save the audio. If saving multiple files,
+                this is treated as a directory and individual files will be saved inside.
+            sampling_rate: Sampling rate for the saved audio. Defaults to the processor's rate.
+            normalize: Whether to normalize audio before saving.
+            batch_prefix: Prefix for batch files when saving multiple audios.
+                
+        Returns:
+            List[str]: Paths to the saved audio files.
+        """
+        if sampling_rate is None:
+            sampling_rate = self.sampling_rate
+        
+        try:
+            import soundfile as sf
+        except ImportError:
+            raise ImportError(
+                "soundfile is required to save audio files. "
+                "Install it with: pip install soundfile"
+            )
+        
+        # Ensure audio is in the right format
+        if isinstance(audio, torch.Tensor):
+            # Convert PyTorch tensor to numpy
+            audio_np = audio.float().detach().cpu().numpy()
+        elif isinstance(audio, np.ndarray):
+            audio_np = audio
+        elif isinstance(audio, list):
+            # Handle list of tensors or arrays
+            if all(isinstance(a, torch.Tensor) for a in audio):
+                audio_np = [a.float().detach().cpu().numpy() for a in audio]
+            else:
+                audio_np = audio
+        else:
+            raise ValueError(f"Unsupported audio type: {type(audio)}")
+        
+        saved_paths = []
+        
+        # Handle based on shape or type
+        if isinstance(audio_np, list):
+            # Multiple separate audios to save
+            output_dir = output_path
+            
+            # Ensure output directory exists
+            os.makedirs(output_dir, exist_ok=True)
+            
+            # Save each audio
+            for i, audio_item in enumerate(audio_np):
+                audio_item = self._prepare_audio_for_save(audio_item, normalize)
+                file_path = os.path.join(output_dir, f"{batch_prefix}{i}.wav")
+                sf.write(file_path, audio_item, sampling_rate)
+                saved_paths.append(file_path)
+                
+        else:
+            # Handle different dimensions
+            if len(audio_np.shape) >= 3:  # (B, C, T) or similar
+                # Get batch size
+                batch_size = audio_np.shape[0]
+                
+                if batch_size > 1:
+                    # Multiple audios in a batch
+                    output_dir = output_path
+                    
+                    # Ensure output directory exists
+                    os.makedirs(output_dir, exist_ok=True)
+                    
+                    # Save each audio in the batch
+                    for i in range(batch_size):
+                        # Extract single audio and remove channel dim if present
+                        single_audio = audio_np[i]
+                        if len(single_audio.shape) > 1:
+                            if single_audio.shape[0] == 1:  # (1, T)
+                                single_audio = single_audio.squeeze(0)
+                        
+                        single_audio = self._prepare_audio_for_save(single_audio, normalize)
+                        file_path = os.path.join(output_dir, f"{batch_prefix}{i}.wav")
+                        sf.write(file_path, single_audio, sampling_rate)
+                        saved_paths.append(file_path)
+                else:
+                    # Single audio with batch and channel dims
+                    audio_item = audio_np.squeeze()  # Remove batch and channel dimensions
+                    audio_item = self._prepare_audio_for_save(audio_item, normalize)
+                    sf.write(output_path, audio_item, sampling_rate)
+                    saved_paths.append(output_path)
+            else:
+                # Single audio without batch dimension
+                audio_item = self._prepare_audio_for_save(audio_np, normalize)
+                sf.write(output_path, audio_item, sampling_rate)
+                saved_paths.append(output_path)
+        
+        return saved_paths
+
+    def _prepare_audio_for_save(self, audio: np.ndarray, normalize: bool) -> np.ndarray:
+        """
+        Prepare audio for saving by ensuring it's the right shape and optionally normalizing.
+        
+        Args:
+            audio: Audio data as numpy array
+            normalize: Whether to normalize audio
+            
+        Returns:
+            np.ndarray: Processed audio ready for saving
+        """
+        # Ensure right dimensionality
+        if len(audio.shape) > 1 and audio.shape[0] == 1:  # (1, T)
+            audio = audio.squeeze(0)
+        
+        # Normalize if requested
+        if normalize:
+            max_val = np.abs(audio).max()
+            if max_val > 0:
+                audio = audio / max_val
+        
+        return audio
+
+
+__all__ = ["VibeVoiceTokenizerProcessor", "AudioNormalizer"]

vvembed/schedule/dpm_solver.py

@@ -0,0 +1,1065 @@
+# Copyright 2024 TSAIL Team and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# DISCLAIMER: This file is strongly influenced by https://github.com/LuChengTHU/dpm-solver
+
+import math
+from typing import List, Optional, Tuple, Union
+
+import numpy as np
+import torch
+
+from diffusers.configuration_utils import ConfigMixin, register_to_config
+from diffusers.utils import deprecate
+from diffusers.utils.torch_utils import randn_tensor
+from diffusers.schedulers.scheduling_utils import KarrasDiffusionSchedulers, SchedulerMixin, SchedulerOutput
+
+def betas_for_alpha_bar(
+    num_diffusion_timesteps,
+    max_beta=0.999,
+    alpha_transform_type="cosine",
+):
+    """
+    Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of
+    (1-beta) over time from t = [0,1].
+
+    Contains a function alpha_bar that takes an argument t and transforms it to the cumulative product of (1-beta) up
+    to that part of the diffusion process.
+
+
+    Args:
+        num_diffusion_timesteps (`int`): the number of betas to produce.
+        max_beta (`float`): the maximum beta to use; use values lower than 1 to
+                     prevent singularities.
+        alpha_transform_type (`str`, *optional*, default to `cosine`): the type of noise schedule for alpha_bar.
+                     Choose from `cosine` or `exp`
+
+    Returns:
+        betas (`np.ndarray`): the betas used by the scheduler to step the model outputs
+    """
+    if alpha_transform_type == "cosine":
+
+        def alpha_bar_fn(t):
+            return math.cos((t + 0.008) / 1.008 * math.pi / 2) ** 2
+            # return math.cos(t * math.pi / 2 * 0.95) ** 2
+
+    elif alpha_transform_type == "exp":
+
+        def alpha_bar_fn(t):
+            return math.exp(t * -12.0)
+
+    elif alpha_transform_type == "cauchy":
+        # µ + γ tan (π (0.5 - x))  γ = 1, µ = 3
+        # alpha^2 = 1-1/(exp(λ)+1)
+        def alpha_bar_fn(t, gamma=1, mu=3):
+            snr = mu + gamma * math.tan(math.pi * (0.5 - t) * 0.9)
+            return 1 - 1 / (math.exp(snr) + 1.1)
+
+    elif alpha_transform_type == "laplace":
+        # µ − bsgn(0.5 − t) log(1 − 2|t − 0.5|) µ = 0, b = 1
+        def alpha_bar_fn(t, mu=0, b=1):
+            snr = mu - b * math.copysign(1, 0.5 - t) * math.log(1 - 2 * abs(t - 0.5) * 0.98)
+            return 1 - 1 / (math.exp(snr) + 1.02)
+
+    else:
+        raise ValueError(f"Unsupported alpha_transform_type: {alpha_transform_type}")
+
+    betas = []
+    for i in range(num_diffusion_timesteps):
+        t1 = i / num_diffusion_timesteps
+        t2 = (i + 1) / num_diffusion_timesteps
+        betas.append(min(1 - alpha_bar_fn(t2) / alpha_bar_fn(t1), max_beta))
+    return torch.tensor(betas, dtype=torch.float32)
+
+
+# Copied from diffusers.schedulers.scheduling_ddim.rescale_zero_terminal_snr
+def rescale_zero_terminal_snr(betas):
+    """
+    Rescales betas to have zero terminal SNR Based on https://arxiv.org/pdf/2305.08891.pdf (Algorithm 1)
+
+
+    Args:
+        betas (`torch.Tensor`):
+            the betas that the scheduler is being initialized with.
+
+    Returns:
+        `torch.Tensor`: rescaled betas with zero terminal SNR
+    """
+    # Convert betas to alphas_bar_sqrt
+    alphas = 1.0 - betas
+    alphas_cumprod = torch.cumprod(alphas, dim=0)
+    alphas_bar_sqrt = alphas_cumprod.sqrt()
+
+    # Store old values.
+    alphas_bar_sqrt_0 = alphas_bar_sqrt[0].clone()
+    alphas_bar_sqrt_T = alphas_bar_sqrt[-1].clone()
+
+    # Shift so the last timestep is zero.
+    alphas_bar_sqrt -= alphas_bar_sqrt_T
+
+    # Scale so the first timestep is back to the old value.
+    alphas_bar_sqrt *= alphas_bar_sqrt_0 / (alphas_bar_sqrt_0 - alphas_bar_sqrt_T)
+
+    # Convert alphas_bar_sqrt to betas
+    alphas_bar = alphas_bar_sqrt**2  # Revert sqrt
+    alphas = alphas_bar[1:] / alphas_bar[:-1]  # Revert cumprod
+    alphas = torch.cat([alphas_bar[0:1], alphas])
+    betas = 1 - alphas
+
+    return betas
+
+class DPMSolverMultistepScheduler(SchedulerMixin, ConfigMixin):
+    """
+    `DPMSolverMultistepScheduler` is a fast dedicated high-order solver for diffusion ODEs.
+
+    This model inherits from [`SchedulerMixin`] and [`ConfigMixin`]. Check the superclass documentation for the generic
+    methods the library implements for all schedulers such as loading and saving.
+
+    Args:
+        num_train_timesteps (`int`, defaults to 1000):
+            The number of diffusion steps to train the model.
+        beta_start (`float`, defaults to 0.0001):
+            The starting `beta` value of inference.
+        beta_end (`float`, defaults to 0.02):
+            The final `beta` value.
+        beta_schedule (`str`, defaults to `"linear"`):
+            The beta schedule, a mapping from a beta range to a sequence of betas for stepping the model. Choose from
+            `linear`, `scaled_linear`, or `squaredcos_cap_v2`.
+        trained_betas (`np.ndarray`, *optional*):
+            Pass an array of betas directly to the constructor to bypass `beta_start` and `beta_end`.
+        solver_order (`int`, defaults to 2):
+            The DPMSolver order which can be `1` or `2` or `3`. It is recommended to use `solver_order=2` for guided
+            sampling, and `solver_order=3` for unconditional sampling.
+        prediction_type (`str`, defaults to `epsilon`, *optional*):
+            Prediction type of the scheduler function; can be `epsilon` (predicts the noise of the diffusion process),
+            `sample` (directly predicts the noisy sample`) or `v_prediction` (see section 2.4 of [Imagen
+            Video](https://imagen.research.google/video/paper.pdf) paper).
+        thresholding (`bool`, defaults to `False`):
+            Whether to use the "dynamic thresholding" method. This is unsuitable for latent-space diffusion models such
+            as Stable Diffusion.
+        dynamic_thresholding_ratio (`float`, defaults to 0.995):
+            The ratio for the dynamic thresholding method. Valid only when `thresholding=True`.
+        sample_max_value (`float`, defaults to 1.0):
+            The threshold value for dynamic thresholding. Valid only when `thresholding=True` and
+            `algorithm_type="dpmsolver++"`.
+        algorithm_type (`str`, defaults to `dpmsolver++`):
+            Algorithm type for the solver; can be `dpmsolver`, `dpmsolver++`, `sde-dpmsolver` or `sde-dpmsolver++`. The
+            `dpmsolver` type implements the algorithms in the [DPMSolver](https://huggingface.co/papers/2206.00927)
+            paper, and the `dpmsolver++` type implements the algorithms in the
+            [DPMSolver++](https://huggingface.co/papers/2211.01095) paper. It is recommended to use `dpmsolver++` or
+            `sde-dpmsolver++` with `solver_order=2` for guided sampling like in Stable Diffusion.
+        solver_type (`str`, defaults to `midpoint`):
+            Solver type for the second-order solver; can be `midpoint` or `heun`. The solver type slightly affects the
+            sample quality, especially for a small number of steps. It is recommended to use `midpoint` solvers.
+        lower_order_final (`bool`, defaults to `True`):
+            Whether to use lower-order solvers in the final steps. Only valid for < 15 inference steps. This can
+            stabilize the sampling of DPMSolver for steps < 15, especially for steps <= 10.
+        euler_at_final (`bool`, defaults to `False`):
+            Whether to use Euler's method in the final step. It is a trade-off between numerical stability and detail
+            richness. This can stabilize the sampling of the SDE variant of DPMSolver for small number of inference
+            steps, but sometimes may result in blurring.
+        use_karras_sigmas (`bool`, *optional*, defaults to `False`):
+            Whether to use Karras sigmas for step sizes in the noise schedule during the sampling process. If `True`,
+            the sigmas are determined according to a sequence of noise levels {σi}.
+        use_lu_lambdas (`bool`, *optional*, defaults to `False`):
+            Whether to use the uniform-logSNR for step sizes proposed by Lu's DPM-Solver in the noise schedule during
+            the sampling process. If `True`, the sigmas and time steps are determined according to a sequence of
+            `lambda(t)`.
+        final_sigmas_type (`str`, defaults to `"zero"`):
+            The final `sigma` value for the noise schedule during the sampling process. If `"sigma_min"`, the final
+            sigma is the same as the last sigma in the training schedule. If `zero`, the final sigma is set to 0.
+        lambda_min_clipped (`float`, defaults to `-inf`):
+            Clipping threshold for the minimum value of `lambda(t)` for numerical stability. This is critical for the
+            cosine (`squaredcos_cap_v2`) noise schedule.
+        variance_type (`str`, *optional*):
+            Set to "learned" or "learned_range" for diffusion models that predict variance. If set, the model's output
+            contains the predicted Gaussian variance.
+        timestep_spacing (`str`, defaults to `"linspace"`):
+            The way the timesteps should be scaled. Refer to Table 2 of the [Common Diffusion Noise Schedules and
+            Sample Steps are Flawed](https://huggingface.co/papers/2305.08891) for more information.
+        steps_offset (`int`, defaults to 0):
+            An offset added to the inference steps, as required by some model families.
+        rescale_betas_zero_snr (`bool`, defaults to `False`):
+            Whether to rescale the betas to have zero terminal SNR. This enables the model to generate very bright and
+            dark samples instead of limiting it to samples with medium brightness. Loosely related to
+            [`--offset_noise`](https://github.com/huggingface/diffusers/blob/74fd735eb073eb1d774b1ab4154a0876eb82f055/examples/dreambooth/train_dreambooth.py#L506).
+    """
+
+    _compatibles = [e.name for e in KarrasDiffusionSchedulers]
+    order = 1
+
+    @register_to_config
+    def __init__(
+        self,
+        num_train_timesteps: int = 1000,
+        beta_start: float = 0.0001,
+        beta_end: float = 0.02,
+        beta_schedule: str = "linear",
+        trained_betas: Optional[Union[np.ndarray, List[float]]] = None,
+        solver_order: int = 2,
+        prediction_type: str = "epsilon",
+        thresholding: bool = False,
+        dynamic_thresholding_ratio: float = 0.995,
+        sample_max_value: float = 1.0,
+        algorithm_type: str = "dpmsolver++",
+        solver_type: str = "midpoint",
+        lower_order_final: bool = True,
+        euler_at_final: bool = False,
+        use_karras_sigmas: Optional[bool] = False,
+        use_lu_lambdas: Optional[bool] = False,
+        final_sigmas_type: Optional[str] = "zero",  # "zero", "sigma_min"
+        lambda_min_clipped: float = -float("inf"),
+        variance_type: Optional[str] = None,
+        timestep_spacing: str = "linspace",
+        steps_offset: int = 0,
+        rescale_betas_zero_snr: bool = False,
+    ):
+        if algorithm_type in ["dpmsolver", "sde-dpmsolver"]:
+            deprecation_message = f"algorithm_type {algorithm_type} is deprecated and will be removed in a future version. Choose from `dpmsolver++` or `sde-dpmsolver++` instead"
+            deprecate("algorithm_types dpmsolver and sde-dpmsolver", "1.0.0", deprecation_message)
+
+        if trained_betas is not None:
+            self.betas = torch.tensor(trained_betas, dtype=torch.float32)
+        elif beta_schedule == "linear":
+            self.betas = torch.linspace(beta_start, beta_end, num_train_timesteps, dtype=torch.float32)
+        elif beta_schedule == "scaled_linear":
+            # this schedule is very specific to the latent diffusion model.
+            self.betas = torch.linspace(beta_start**0.5, beta_end**0.5, num_train_timesteps, dtype=torch.float32) ** 2
+        elif beta_schedule == "squaredcos_cap_v2" or beta_schedule == "cosine":
+            # Glide cosine schedule
+            self.betas = betas_for_alpha_bar(num_train_timesteps, alpha_transform_type="cosine")
+        elif beta_schedule == "cauchy":
+            self.betas = betas_for_alpha_bar(num_train_timesteps, alpha_transform_type="cauchy")
+        elif beta_schedule == "laplace":
+            self.betas = betas_for_alpha_bar(num_train_timesteps, alpha_transform_type="laplace")
+        else:
+            raise NotImplementedError(f"{beta_schedule} is not implemented for {self.__class__}")
+
+        if rescale_betas_zero_snr:
+            self.betas = rescale_zero_terminal_snr(self.betas)
+
+        self.alphas = 1.0 - self.betas
+        self.alphas_cumprod = torch.cumprod(self.alphas, dim=0)
+
+        if rescale_betas_zero_snr:
+            # Close to 0 without being 0 so first sigma is not inf
+            # FP16 smallest positive subnormal works well here
+            self.alphas_cumprod[-1] = 2**-24
+
+        # Currently we only support VP-type noise schedule
+        self.alpha_t = torch.sqrt(self.alphas_cumprod)
+        self.sigma_t = torch.sqrt(1 - self.alphas_cumprod)
+        self.lambda_t = torch.log(self.alpha_t) - torch.log(self.sigma_t)
+        self.sigmas = ((1 - self.alphas_cumprod) / self.alphas_cumprod) ** 0.5
+
+        # standard deviation of the initial noise distribution
+        self.init_noise_sigma = 1.0
+
+        # settings for DPM-Solver
+        if algorithm_type not in ["dpmsolver", "dpmsolver++", "sde-dpmsolver", "sde-dpmsolver++"]:
+            if algorithm_type == "deis":
+                self.register_to_config(algorithm_type="dpmsolver++")
+            else:
+                raise NotImplementedError(f"{algorithm_type} is not implemented for {self.__class__}")
+
+        if solver_type not in ["midpoint", "heun"]:
+            if solver_type in ["logrho", "bh1", "bh2"]:
+                self.register_to_config(solver_type="midpoint")
+            else:
+                raise NotImplementedError(f"{solver_type} is not implemented for {self.__class__}")
+
+        if algorithm_type not in ["dpmsolver++", "sde-dpmsolver++"] and final_sigmas_type == "zero":
+            raise ValueError(
+                f"`final_sigmas_type` {final_sigmas_type} is not supported for `algorithm_type` {algorithm_type}. Please choose `sigma_min` instead."
+            )
+
+        # setable values
+        self.num_inference_steps = None
+        timesteps = np.linspace(0, num_train_timesteps - 1, num_train_timesteps, dtype=np.float32)[::-1].copy()
+        self.timesteps = torch.from_numpy(timesteps)
+        self.model_outputs = [None] * solver_order
+        self.lower_order_nums = 0
+        self._step_index = None
+        self._begin_index = None
+        self.sigmas = self.sigmas.to("cpu")  # to avoid too much CPU/GPU communication
+
+    @property
+    def step_index(self):
+        """
+        The index counter for current timestep. It will increase 1 after each scheduler step.
+        """
+        return self._step_index
+
+    @property
+    def begin_index(self):
+        """
+        The index for the first timestep. It should be set from pipeline with `set_begin_index` method.
+        """
+        return self._begin_index
+
+    def set_begin_index(self, begin_index: int = 0):
+        """
+        Sets the begin index for the scheduler. This function should be run from pipeline before the inference.
+
+        Args:
+            begin_index (`int`):
+                The begin index for the scheduler.
+        """
+        self._begin_index = begin_index
+
+    def set_timesteps(
+        self,
+        num_inference_steps: int = None,
+        device: Union[str, torch.device] = None,
+        timesteps: Optional[List[int]] = None,
+    ):
+        """
+        Sets the discrete timesteps used for the diffusion chain (to be run before inference).
+
+        Args:
+            num_inference_steps (`int`):
+                The number of diffusion steps used when generating samples with a pre-trained model.
+            device (`str` or `torch.device`, *optional*):
+                The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+            timesteps (`List[int]`, *optional*):
+                Custom timesteps used to support arbitrary timesteps schedule. If `None`, timesteps will be generated
+                based on the `timestep_spacing` attribute. If `timesteps` is passed, `num_inference_steps` and `sigmas`
+                must be `None`, and `timestep_spacing` attribute will be ignored.
+        """
+        if num_inference_steps is None and timesteps is None:
+            raise ValueError("Must pass exactly one of `num_inference_steps` or `timesteps`.")
+        if num_inference_steps is not None and timesteps is not None:
+            raise ValueError("Can only pass one of `num_inference_steps` or `custom_timesteps`.")
+        if timesteps is not None and self.config.use_karras_sigmas:
+            raise ValueError("Cannot use `timesteps` with `config.use_karras_sigmas = True`")
+        if timesteps is not None and self.config.use_lu_lambdas:
+            raise ValueError("Cannot use `timesteps` with `config.use_lu_lambdas = True`")
+
+        if timesteps is not None:
+            timesteps = np.array(timesteps).astype(np.int64)
+        else:
+            # Clipping the minimum of all lambda(t) for numerical stability.
+            # This is critical for cosine (squaredcos_cap_v2) noise schedule.
+            clipped_idx = torch.searchsorted(torch.flip(self.lambda_t, [0]), self.config.lambda_min_clipped)
+            last_timestep = ((self.config.num_train_timesteps - clipped_idx).numpy()).item()
+
+            # "linspace", "leading", "trailing" corresponds to annotation of Table 2. of https://arxiv.org/abs/2305.08891
+            if self.config.timestep_spacing == "linspace":
+                timesteps = (
+                    np.linspace(0, last_timestep - 1, num_inference_steps + 1)
+                    .round()[::-1][:-1]
+                    .copy()
+                    .astype(np.int64)
+                )
+            elif self.config.timestep_spacing == "leading":
+                step_ratio = last_timestep // (num_inference_steps + 1)
+                # creates integer timesteps by multiplying by ratio
+                # casting to int to avoid issues when num_inference_step is power of 3
+                timesteps = (
+                    (np.arange(0, num_inference_steps + 1) * step_ratio).round()[::-1][:-1].copy().astype(np.int64)
+                )
+                timesteps += self.config.steps_offset
+            elif self.config.timestep_spacing == "trailing":
+                step_ratio = self.config.num_train_timesteps / num_inference_steps
+                # creates integer timesteps by multiplying by ratio
+                # casting to int to avoid issues when num_inference_step is power of 3
+                timesteps = np.arange(last_timestep, 0, -step_ratio).round().copy().astype(np.int64)
+                timesteps -= 1
+            else:
+                raise ValueError(
+                    f"{self.config.timestep_spacing} is not supported. Please make sure to choose one of 'linspace', 'leading' or 'trailing'."
+                )
+
+        sigmas = np.array(((1 - self.alphas_cumprod) / self.alphas_cumprod) ** 0.5)
+        log_sigmas = np.log(sigmas)
+
+        if self.config.use_karras_sigmas:
+            sigmas = np.flip(sigmas).copy()
+            sigmas = self._convert_to_karras(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
+            timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas]).round()
+        elif self.config.use_lu_lambdas:
+            lambdas = np.flip(log_sigmas.copy())
+            lambdas = self._convert_to_lu(in_lambdas=lambdas, num_inference_steps=num_inference_steps)
+            sigmas = np.exp(lambdas)
+            timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas]).round()
+        else:
+            sigmas = np.interp(timesteps, np.arange(0, len(sigmas)), sigmas)
+
+        if self.config.final_sigmas_type == "sigma_min":
+            sigma_last = ((1 - self.alphas_cumprod[0]) / self.alphas_cumprod[0]) ** 0.5
+        elif self.config.final_sigmas_type == "zero":
+            sigma_last = 0
+        else:
+            raise ValueError(
+                f"`final_sigmas_type` must be one of 'zero', or 'sigma_min', but got {self.config.final_sigmas_type}"
+            )
+
+        sigmas = np.concatenate([sigmas, [sigma_last]]).astype(np.float32)
+
+        self.sigmas = torch.from_numpy(sigmas)
+        self.timesteps = torch.from_numpy(timesteps).to(device=device, dtype=torch.int64)
+
+        self.num_inference_steps = len(timesteps)
+
+        self.model_outputs = [
+            None,
+        ] * self.config.solver_order
+        self.lower_order_nums = 0
+
+        # add an index counter for schedulers that allow duplicated timesteps
+        self._step_index = None
+        self._begin_index = None
+        self.sigmas = self.sigmas.to("cpu")  # to avoid too much CPU/GPU communication
+
+    # Copied from diffusers.schedulers.scheduling_ddpm.DDPMScheduler._threshold_sample
+    def _threshold_sample(self, sample: torch.Tensor) -> torch.Tensor:
+        """
+        "Dynamic thresholding: At each sampling step we set s to a certain percentile absolute pixel value in xt0 (the
+        prediction of x_0 at timestep t), and if s > 1, then we threshold xt0 to the range [-s, s] and then divide by
+        s. Dynamic thresholding pushes saturated pixels (those near -1 and 1) inwards, thereby actively preventing
+        pixels from saturation at each step. We find that dynamic thresholding results in significantly better
+        photorealism as well as better image-text alignment, especially when using very large guidance weights."
+
+        https://arxiv.org/abs/2205.11487
+        """
+        dtype = sample.dtype
+        batch_size, channels, *remaining_dims = sample.shape
+
+        if dtype not in (torch.float32, torch.float64):
+            sample = sample.float()  # upcast for quantile calculation, and clamp not implemented for cpu half
+
+        # Flatten sample for doing quantile calculation along each image
+        sample = sample.reshape(batch_size, channels * np.prod(remaining_dims))
+
+        abs_sample = sample.abs()  # "a certain percentile absolute pixel value"
+
+        s = torch.quantile(abs_sample, self.config.dynamic_thresholding_ratio, dim=1)
+        s = torch.clamp(
+            s, min=1, max=self.config.sample_max_value
+        )  # When clamped to min=1, equivalent to standard clipping to [-1, 1]
+        s = s.unsqueeze(1)  # (batch_size, 1) because clamp will broadcast along dim=0
+        sample = torch.clamp(sample, -s, s) / s  # "we threshold xt0 to the range [-s, s] and then divide by s"
+
+        sample = sample.reshape(batch_size, channels, *remaining_dims)
+        sample = sample.to(dtype)
+
+        return sample
+
+    # Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._sigma_to_t
+    def _sigma_to_t(self, sigma, log_sigmas):
+        # get log sigma
+        log_sigma = np.log(np.maximum(sigma, 1e-10))
+
+        # get distribution
+        dists = log_sigma - log_sigmas[:, np.newaxis]
+
+        # get sigmas range
+        low_idx = np.cumsum((dists >= 0), axis=0).argmax(axis=0).clip(max=log_sigmas.shape[0] - 2)
+        high_idx = low_idx + 1
+
+        low = log_sigmas[low_idx]
+        high = log_sigmas[high_idx]
+
+        # interpolate sigmas
+        w = (low - log_sigma) / (low - high)
+        w = np.clip(w, 0, 1)
+
+        # transform interpolation to time range
+        t = (1 - w) * low_idx + w * high_idx
+        t = t.reshape(sigma.shape)
+        return t
+
+    def _sigma_to_alpha_sigma_t(self, sigma):
+        alpha_t = 1 / ((sigma**2 + 1) ** 0.5)
+        sigma_t = sigma * alpha_t
+
+        return alpha_t, sigma_t
+
+    # Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_karras
+    def _convert_to_karras(self, in_sigmas: torch.Tensor, num_inference_steps) -> torch.Tensor:
+        """Constructs the noise schedule of Karras et al. (2022)."""
+
+        # Hack to make sure that other schedulers which copy this function don't break
+        # TODO: Add this logic to the other schedulers
+        if hasattr(self.config, "sigma_min"):
+            sigma_min = self.config.sigma_min
+        else:
+            sigma_min = None
+
+        if hasattr(self.config, "sigma_max"):
+            sigma_max = self.config.sigma_max
+        else:
+            sigma_max = None
+
+        sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
+        sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()
+
+        rho = 7.0  # 7.0 is the value used in the paper
+        ramp = np.linspace(0, 1, num_inference_steps)
+        min_inv_rho = sigma_min ** (1 / rho)
+        max_inv_rho = sigma_max ** (1 / rho)
+        sigmas = (max_inv_rho + ramp * (min_inv_rho - max_inv_rho)) ** rho
+        return sigmas
+
+    def _convert_to_lu(self, in_lambdas: torch.Tensor, num_inference_steps) -> torch.Tensor:
+        """Constructs the noise schedule of Lu et al. (2022)."""
+
+        lambda_min: float = in_lambdas[-1].item()
+        lambda_max: float = in_lambdas[0].item()
+
+        rho = 1.0  # 1.0 is the value used in the paper
+        ramp = np.linspace(0, 1, num_inference_steps)
+        min_inv_rho = lambda_min ** (1 / rho)
+        max_inv_rho = lambda_max ** (1 / rho)
+        lambdas = (max_inv_rho + ramp * (min_inv_rho - max_inv_rho)) ** rho
+        return lambdas
+
+    def convert_model_output(
+        self,
+        model_output: torch.Tensor,
+        *args,
+        sample: torch.Tensor = None,
+        **kwargs,
+    ) -> torch.Tensor:
+        """
+        Convert the model output to the corresponding type the DPMSolver/DPMSolver++ algorithm needs. DPM-Solver is
+        designed to discretize an integral of the noise prediction model, and DPM-Solver++ is designed to discretize an
+        integral of the data prediction model.
+
+        <Tip>
+
+        The algorithm and model type are decoupled. You can use either DPMSolver or DPMSolver++ for both noise
+        prediction and data prediction models.
+
+        </Tip>
+
+        Args:
+            model_output (`torch.Tensor`):
+                The direct output from the learned diffusion model.
+            sample (`torch.Tensor`):
+                A current instance of a sample created by the diffusion process.
+
+        Returns:
+            `torch.Tensor`:
+                The converted model output.
+        """
+        timestep = args[0] if len(args) > 0 else kwargs.pop("timestep", None)
+        if sample is None:
+            if len(args) > 1:
+                sample = args[1]
+            else:
+                raise ValueError("missing `sample` as a required keyward argument")
+        if timestep is not None:
+            deprecate(
+                "timesteps",
+                "1.0.0",
+                "Passing `timesteps` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`",
+            )
+
+        # DPM-Solver++ needs to solve an integral of the data prediction model.
+        if self.config.algorithm_type in ["dpmsolver++", "sde-dpmsolver++"]:
+            if self.config.prediction_type == "epsilon":
+                # DPM-Solver and DPM-Solver++ only need the "mean" output.
+                if self.config.variance_type in ["learned", "learned_range"]:
+                    model_output = model_output[:, :3]
+                sigma = self.sigmas[self.step_index]
+                alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma)
+                x0_pred = (sample - sigma_t * model_output) / alpha_t
+            elif self.config.prediction_type == "sample":
+                x0_pred = model_output
+            elif self.config.prediction_type == "v_prediction":
+                sigma = self.sigmas[self.step_index]
+                alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma)
+                x0_pred = alpha_t * sample - sigma_t * model_output
+            else:
+                raise ValueError(
+                    f"prediction_type given as {self.config.prediction_type} must be one of `epsilon`, `sample`, or"
+                    " `v_prediction` for the DPMSolverMultistepScheduler."
+                )
+
+            if self.config.thresholding:
+                x0_pred = self._threshold_sample(x0_pred)
+
+            return x0_pred
+
+        # DPM-Solver needs to solve an integral of the noise prediction model.
+        elif self.config.algorithm_type in ["dpmsolver", "sde-dpmsolver"]:
+            if self.config.prediction_type == "epsilon":
+                # DPM-Solver and DPM-Solver++ only need the "mean" output.
+                if self.config.variance_type in ["learned", "learned_range"]:
+                    epsilon = model_output[:, :3]
+                else:
+                    epsilon = model_output
+            elif self.config.prediction_type == "sample":
+                sigma = self.sigmas[self.step_index]
+                alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma)
+                epsilon = (sample - alpha_t * model_output) / sigma_t
+            elif self.config.prediction_type == "v_prediction":
+                sigma = self.sigmas[self.step_index]
+                alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma)
+                epsilon = alpha_t * model_output + sigma_t * sample
+            else:
+                raise ValueError(
+                    f"prediction_type given as {self.config.prediction_type} must be one of `epsilon`, `sample`, or"
+                    " `v_prediction` for the DPMSolverMultistepScheduler."
+                )
+
+            if self.config.thresholding:
+                sigma = self.sigmas[self.step_index]
+                alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma)
+                x0_pred = (sample - sigma_t * epsilon) / alpha_t
+                x0_pred = self._threshold_sample(x0_pred)
+                epsilon = (sample - alpha_t * x0_pred) / sigma_t
+
+            return epsilon
+
+    def dpm_solver_first_order_update(
+        self,
+        model_output: torch.Tensor,
+        *args,
+        sample: torch.Tensor = None,
+        noise: Optional[torch.Tensor] = None,
+        **kwargs,
+    ) -> torch.Tensor:
+        """
+        One step for the first-order DPMSolver (equivalent to DDIM).
+
+        Args:
+            model_output (`torch.Tensor`):
+                The direct output from the learned diffusion model.
+            sample (`torch.Tensor`):
+                A current instance of a sample created by the diffusion process.
+
+        Returns:
+            `torch.Tensor`:
+                The sample tensor at the previous timestep.
+        """
+        timestep = args[0] if len(args) > 0 else kwargs.pop("timestep", None)
+        prev_timestep = args[1] if len(args) > 1 else kwargs.pop("prev_timestep", None)
+        if sample is None:
+            if len(args) > 2:
+                sample = args[2]
+            else:
+                raise ValueError(" missing `sample` as a required keyward argument")
+        if timestep is not None:
+            deprecate(
+                "timesteps",
+                "1.0.0",
+                "Passing `timesteps` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`",
+            )
+
+        if prev_timestep is not None:
+            deprecate(
+                "prev_timestep",
+                "1.0.0",
+                "Passing `prev_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`",
+            )
+
+        sigma_t, sigma_s = self.sigmas[self.step_index + 1], self.sigmas[self.step_index]
+        alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma_t)
+        alpha_s, sigma_s = self._sigma_to_alpha_sigma_t(sigma_s)
+        lambda_t = torch.log(alpha_t) - torch.log(sigma_t)
+        lambda_s = torch.log(alpha_s) - torch.log(sigma_s)
+
+        h = lambda_t - lambda_s
+        if self.config.algorithm_type == "dpmsolver++":
+            x_t = (sigma_t / sigma_s) * sample - (alpha_t * (torch.exp(-h) - 1.0)) * model_output
+        elif self.config.algorithm_type == "dpmsolver":
+            x_t = (alpha_t / alpha_s) * sample - (sigma_t * (torch.exp(h) - 1.0)) * model_output
+        elif self.config.algorithm_type == "sde-dpmsolver++":
+            assert noise is not None
+            x_t = (
+                (sigma_t / sigma_s * torch.exp(-h)) * sample
+                + (alpha_t * (1 - torch.exp(-2.0 * h))) * model_output
+                + sigma_t * torch.sqrt(1.0 - torch.exp(-2 * h)) * noise
+            )
+        elif self.config.algorithm_type == "sde-dpmsolver":
+            assert noise is not None
+            x_t = (
+                (alpha_t / alpha_s) * sample
+                - 2.0 * (sigma_t * (torch.exp(h) - 1.0)) * model_output
+                + sigma_t * torch.sqrt(torch.exp(2 * h) - 1.0) * noise
+            )
+        return x_t
+
+    def multistep_dpm_solver_second_order_update(
+        self,
+        model_output_list: List[torch.Tensor],
+        *args,
+        sample: torch.Tensor = None,
+        noise: Optional[torch.Tensor] = None,
+        **kwargs,
+    ) -> torch.Tensor:
+        """
+        One step for the second-order multistep DPMSolver.
+
+        Args:
+            model_output_list (`List[torch.Tensor]`):
+                The direct outputs from learned diffusion model at current and latter timesteps.
+            sample (`torch.Tensor`):
+                A current instance of a sample created by the diffusion process.
+
+        Returns:
+            `torch.Tensor`:
+                The sample tensor at the previous timestep.
+        """
+        timestep_list = args[0] if len(args) > 0 else kwargs.pop("timestep_list", None)
+        prev_timestep = args[1] if len(args) > 1 else kwargs.pop("prev_timestep", None)
+        if sample is None:
+            if len(args) > 2:
+                sample = args[2]
+            else:
+                raise ValueError(" missing `sample` as a required keyward argument")
+        if timestep_list is not None:
+            deprecate(
+                "timestep_list",
+                "1.0.0",
+                "Passing `timestep_list` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`",
+            )
+
+        if prev_timestep is not None:
+            deprecate(
+                "prev_timestep",
+                "1.0.0",
+                "Passing `prev_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`",
+            )
+
+        sigma_t, sigma_s0, sigma_s1 = (
+            self.sigmas[self.step_index + 1],
+            self.sigmas[self.step_index],
+            self.sigmas[self.step_index - 1],
+        )
+
+        alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma_t)
+        alpha_s0, sigma_s0 = self._sigma_to_alpha_sigma_t(sigma_s0)
+        alpha_s1, sigma_s1 = self._sigma_to_alpha_sigma_t(sigma_s1)
+
+        lambda_t = torch.log(alpha_t) - torch.log(sigma_t)
+        lambda_s0 = torch.log(alpha_s0) - torch.log(sigma_s0)
+        lambda_s1 = torch.log(alpha_s1) - torch.log(sigma_s1)
+
+        m0, m1 = model_output_list[-1], model_output_list[-2]
+
+        h, h_0 = lambda_t - lambda_s0, lambda_s0 - lambda_s1
+        r0 = h_0 / h
+        D0, D1 = m0, (1.0 / r0) * (m0 - m1)
+        if self.config.algorithm_type == "dpmsolver++":
+            # See https://arxiv.org/abs/2211.01095 for detailed derivations
+            if self.config.solver_type == "midpoint":
+                x_t = (
+                    (sigma_t / sigma_s0) * sample
+                    - (alpha_t * (torch.exp(-h) - 1.0)) * D0
+                    - 0.5 * (alpha_t * (torch.exp(-h) - 1.0)) * D1
+                )
+            elif self.config.solver_type == "heun":
+                x_t = (
+                    (sigma_t / sigma_s0) * sample
+                    - (alpha_t * (torch.exp(-h) - 1.0)) * D0
+                    + (alpha_t * ((torch.exp(-h) - 1.0) / h + 1.0)) * D1
+                )
+        elif self.config.algorithm_type == "dpmsolver":
+            # See https://arxiv.org/abs/2206.00927 for detailed derivations
+            if self.config.solver_type == "midpoint":
+                x_t = (
+                    (alpha_t / alpha_s0) * sample
+                    - (sigma_t * (torch.exp(h) - 1.0)) * D0
+                    - 0.5 * (sigma_t * (torch.exp(h) - 1.0)) * D1
+                )
+            elif self.config.solver_type == "heun":
+                x_t = (
+                    (alpha_t / alpha_s0) * sample
+                    - (sigma_t * (torch.exp(h) - 1.0)) * D0
+                    - (sigma_t * ((torch.exp(h) - 1.0) / h - 1.0)) * D1
+                )
+        elif self.config.algorithm_type == "sde-dpmsolver++":
+            assert noise is not None
+            if self.config.solver_type == "midpoint":
+                x_t = (
+                    (sigma_t / sigma_s0 * torch.exp(-h)) * sample
+                    + (alpha_t * (1 - torch.exp(-2.0 * h))) * D0
+                    + 0.5 * (alpha_t * (1 - torch.exp(-2.0 * h))) * D1
+                    + sigma_t * torch.sqrt(1.0 - torch.exp(-2 * h)) * noise
+                )
+            elif self.config.solver_type == "heun":
+                x_t = (
+                    (sigma_t / sigma_s0 * torch.exp(-h)) * sample
+                    + (alpha_t * (1 - torch.exp(-2.0 * h))) * D0
+                    + (alpha_t * ((1.0 - torch.exp(-2.0 * h)) / (-2.0 * h) + 1.0)) * D1
+                    + sigma_t * torch.sqrt(1.0 - torch.exp(-2 * h)) * noise
+                )
+        elif self.config.algorithm_type == "sde-dpmsolver":
+            assert noise is not None
+            if self.config.solver_type == "midpoint":
+                x_t = (
+                    (alpha_t / alpha_s0) * sample
+                    - 2.0 * (sigma_t * (torch.exp(h) - 1.0)) * D0
+                    - (sigma_t * (torch.exp(h) - 1.0)) * D1
+                    + sigma_t * torch.sqrt(torch.exp(2 * h) - 1.0) * noise
+                )
+            elif self.config.solver_type == "heun":
+                x_t = (
+                    (alpha_t / alpha_s0) * sample
+                    - 2.0 * (sigma_t * (torch.exp(h) - 1.0)) * D0
+                    - 2.0 * (sigma_t * ((torch.exp(h) - 1.0) / h - 1.0)) * D1
+                    + sigma_t * torch.sqrt(torch.exp(2 * h) - 1.0) * noise
+                )
+        return x_t
+
+    def multistep_dpm_solver_third_order_update(
+        self,
+        model_output_list: List[torch.Tensor],
+        *args,
+        sample: torch.Tensor = None,
+        **kwargs,
+    ) -> torch.Tensor:
+        """
+        One step for the third-order multistep DPMSolver.
+
+        Args:
+            model_output_list (`List[torch.Tensor]`):
+                The direct outputs from learned diffusion model at current and latter timesteps.
+            sample (`torch.Tensor`):
+                A current instance of a sample created by diffusion process.
+
+        Returns:
+            `torch.Tensor`:
+                The sample tensor at the previous timestep.
+        """
+
+        timestep_list = args[0] if len(args) > 0 else kwargs.pop("timestep_list", None)
+        prev_timestep = args[1] if len(args) > 1 else kwargs.pop("prev_timestep", None)
+        if sample is None:
+            if len(args) > 2:
+                sample = args[2]
+            else:
+                raise ValueError(" missing`sample` as a required keyward argument")
+        if timestep_list is not None:
+            deprecate(
+                "timestep_list",
+                "1.0.0",
+                "Passing `timestep_list` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`",
+            )
+
+        if prev_timestep is not None:
+            deprecate(
+                "prev_timestep",
+                "1.0.0",
+                "Passing `prev_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`",
+            )
+
+        sigma_t, sigma_s0, sigma_s1, sigma_s2 = (
+            self.sigmas[self.step_index + 1],
+            self.sigmas[self.step_index],
+            self.sigmas[self.step_index - 1],
+            self.sigmas[self.step_index - 2],
+        )
+
+        alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma_t)
+        alpha_s0, sigma_s0 = self._sigma_to_alpha_sigma_t(sigma_s0)
+        alpha_s1, sigma_s1 = self._sigma_to_alpha_sigma_t(sigma_s1)
+        alpha_s2, sigma_s2 = self._sigma_to_alpha_sigma_t(sigma_s2)
+
+        lambda_t = torch.log(alpha_t) - torch.log(sigma_t)
+        lambda_s0 = torch.log(alpha_s0) - torch.log(sigma_s0)
+        lambda_s1 = torch.log(alpha_s1) - torch.log(sigma_s1)
+        lambda_s2 = torch.log(alpha_s2) - torch.log(sigma_s2)
+
+        m0, m1, m2 = model_output_list[-1], model_output_list[-2], model_output_list[-3]
+
+        h, h_0, h_1 = lambda_t - lambda_s0, lambda_s0 - lambda_s1, lambda_s1 - lambda_s2
+        r0, r1 = h_0 / h, h_1 / h
+        D0 = m0
+        D1_0, D1_1 = (1.0 / r0) * (m0 - m1), (1.0 / r1) * (m1 - m2)
+        D1 = D1_0 + (r0 / (r0 + r1)) * (D1_0 - D1_1)
+        D2 = (1.0 / (r0 + r1)) * (D1_0 - D1_1)
+        if self.config.algorithm_type == "dpmsolver++":
+            # See https://arxiv.org/abs/2206.00927 for detailed derivations
+            x_t = (
+                (sigma_t / sigma_s0) * sample
+                - (alpha_t * (torch.exp(-h) - 1.0)) * D0
+                + (alpha_t * ((torch.exp(-h) - 1.0) / h + 1.0)) * D1
+                - (alpha_t * ((torch.exp(-h) - 1.0 + h) / h**2 - 0.5)) * D2
+            )
+        elif self.config.algorithm_type == "dpmsolver":
+            # See https://arxiv.org/abs/2206.00927 for detailed derivations
+            x_t = (
+                (alpha_t / alpha_s0) * sample
+                - (sigma_t * (torch.exp(h) - 1.0)) * D0
+                - (sigma_t * ((torch.exp(h) - 1.0) / h - 1.0)) * D1
+                - (sigma_t * ((torch.exp(h) - 1.0 - h) / h**2 - 0.5)) * D2
+            )
+        return x_t
+
+    def index_for_timestep(self, timestep, schedule_timesteps=None):
+        if schedule_timesteps is None:
+            schedule_timesteps = self.timesteps
+
+        index_candidates = (schedule_timesteps == timestep).nonzero()
+
+        if len(index_candidates) == 0:
+            step_index = len(self.timesteps) - 1
+        # The sigma index that is taken for the **very** first `step`
+        # is always the second index (or the last index if there is only 1)
+        # This way we can ensure we don't accidentally skip a sigma in
+        # case we start in the middle of the denoising schedule (e.g. for image-to-image)
+        elif len(index_candidates) > 1:
+            step_index = index_candidates[1].item()
+        else:
+            step_index = index_candidates[0].item()
+
+        return step_index
+
+    def _init_step_index(self, timestep):
+        """
+        Initialize the step_index counter for the scheduler.
+        """
+
+        if self.begin_index is None:
+            if isinstance(timestep, torch.Tensor):
+                timestep = timestep.to(self.timesteps.device)
+            self._step_index = self.index_for_timestep(timestep)
+        else:
+            self._step_index = self._begin_index
+
+    def step(
+        self,
+        model_output: torch.Tensor,
+        timestep: int,
+        sample: torch.Tensor,
+        generator=None,
+        variance_noise: Optional[torch.Tensor] = None,
+        return_dict: bool = True,
+    ) -> Union[SchedulerOutput, Tuple]:
+        """
+        Predict the sample from the previous timestep by reversing the SDE. This function propagates the sample with
+        the multistep DPMSolver.
+
+        Args:
+            model_output (`torch.Tensor`):
+                The direct output from learned diffusion model.
+            timestep (`int`):
+                The current discrete timestep in the diffusion chain.
+            sample (`torch.Tensor`):
+                A current instance of a sample created by the diffusion process.
+            generator (`torch.Generator`, *optional*):
+                A random number generator.
+            variance_noise (`torch.Tensor`):
+                Alternative to generating noise with `generator` by directly providing the noise for the variance
+                itself. Useful for methods such as [`LEdits++`].
+            return_dict (`bool`):
+                Whether or not to return a [`~schedulers.scheduling_utils.SchedulerOutput`] or `tuple`.
+
+        Returns:
+            [`~schedulers.scheduling_utils.SchedulerOutput`] or `tuple`:
+                If return_dict is `True`, [`~schedulers.scheduling_utils.SchedulerOutput`] is returned, otherwise a
+                tuple is returned where the first element is the sample tensor.
+
+        """
+        if self.num_inference_steps is None:
+            raise ValueError(
+                "Number of inference steps is 'None', you need to run 'set_timesteps' after creating the scheduler"
+            )
+
+        if self.step_index is None:
+            self._init_step_index(timestep)
+
+        # Improve numerical stability for small number of steps
+        lower_order_final = (self.step_index == len(self.timesteps) - 1) and (
+            self.config.euler_at_final
+            or (self.config.lower_order_final and len(self.timesteps) < 15)
+            or self.config.final_sigmas_type == "zero"
+        )
+        lower_order_second = (
+            (self.step_index == len(self.timesteps) - 2) and self.config.lower_order_final and len(self.timesteps) < 15
+        )
+
+        model_output = self.convert_model_output(model_output, sample=sample)
+        for i in range(self.config.solver_order - 1):
+            self.model_outputs[i] = self.model_outputs[i + 1]
+        self.model_outputs[-1] = model_output
+
+        # Upcast to avoid precision issues when computing prev_sample
+        sample = sample.to(torch.float32)
+        if self.config.algorithm_type in ["sde-dpmsolver", "sde-dpmsolver++"] and variance_noise is None:
+            noise = randn_tensor(
+                model_output.shape, generator=generator, device=model_output.device, dtype=torch.float32
+            )
+        elif self.config.algorithm_type in ["sde-dpmsolver", "sde-dpmsolver++"]:
+            noise = variance_noise.to(device=model_output.device, dtype=torch.float32)
+        else:
+            noise = None
+
+        if self.config.solver_order == 1 or self.lower_order_nums < 1 or lower_order_final:
+            prev_sample = self.dpm_solver_first_order_update(model_output, sample=sample, noise=noise)
+        elif self.config.solver_order == 2 or self.lower_order_nums < 2 or lower_order_second:
+            prev_sample = self.multistep_dpm_solver_second_order_update(self.model_outputs, sample=sample, noise=noise)
+        else:
+            prev_sample = self.multistep_dpm_solver_third_order_update(self.model_outputs, sample=sample)
+
+        if self.lower_order_nums < self.config.solver_order:
+            self.lower_order_nums += 1
+
+        # Cast sample back to expected dtype
+        prev_sample = prev_sample.to(model_output.dtype)
+
+        # upon completion increase step index by one
+        self._step_index += 1
+
+        if not return_dict:
+            return (prev_sample,)
+
+        return SchedulerOutput(prev_sample=prev_sample)
+
+    def add_noise(
+        self,
+        original_samples: torch.Tensor,
+        noise: torch.Tensor,
+        timesteps: torch.IntTensor,
+    ) -> torch.Tensor:
+        # Make sure sigmas and timesteps have the same device and dtype as original_samples
+        # alpha_t = self.alpha_t.to(device=original_samples.device, dtype=original_samples.dtype)
+        # sigma_t = self.sigma_t.to(device=original_samples.device, dtype=original_samples.dtype)
+        alpha_t = self.alpha_t.to(original_samples.device).to(original_samples.dtype)
+        sigma_t = self.sigma_t.to(original_samples.device).to(original_samples.dtype)
+        timesteps = timesteps.to(original_samples.device)
+        alpha_t = alpha_t[timesteps].flatten()
+        while len(alpha_t.shape) < len(original_samples.shape):
+            alpha_t = alpha_t.unsqueeze(-1)
+
+        sigma_t = sigma_t[timesteps].flatten()
+        while len(sigma_t.shape) < len(original_samples.shape):
+            sigma_t = sigma_t.unsqueeze(-1)
+        noisy_samples = alpha_t * original_samples + sigma_t * noise
+        return noisy_samples
+    
+    def get_velocity(self, original_samples: torch.Tensor, noise: torch.Tensor, timesteps: torch.IntTensor) -> torch.Tensor:
+        # alpha_t = self.alpha_t.to(device=original_samples.device, dtype=original_samples.dtype)
+        # sigma_t = self.sigma_t.to(device=original_samples.device, dtype=original_samples.dtype)
+        alpha_t = self.alpha_t.to(original_samples.device).to(original_samples.dtype)
+        sigma_t = self.sigma_t.to(original_samples.device).to(original_samples.dtype)
+
+        timesteps = timesteps.to(original_samples.device)
+        alpha_t = alpha_t[timesteps].flatten()
+        while len(alpha_t.shape) < len(original_samples.shape):
+            alpha_t = alpha_t.unsqueeze(-1)
+
+        sigma_t = sigma_t[timesteps].flatten()
+        while len(sigma_t.shape) < len(original_samples.shape):
+            sigma_t = sigma_t.unsqueeze(-1)
+
+        velocity = alpha_t * noise - sigma_t * original_samples
+        return velocity
+
+    def __len__(self):
+        return self.config.num_train_timesteps

vvembed/schedule/timestep_sampler.py

@@ -0,0 +1,19 @@
+import math
+import torch
+
+
+class UniformSampler:
+    def __init__(self, timesteps = 1000):
+        self.timesteps = timesteps
+    def sample(self, batch_size, device):
+        return torch.randint(0, self.timesteps, (batch_size,), device=device)
+    
+class LogitNormalSampler:
+    def __init__(self, timesteps = 1000, m = 0, s = 1):
+        self.timesteps = timesteps
+        timesteps = torch.linspace(0, 1, timesteps)
+        logit = torch.log(timesteps / (1 - timesteps))
+        self.prob = torch.exp(-0.5 * (logit - m) ** 2 / s ** 2) / (s * math.sqrt(2 * math.pi))
+    def sample(self, batch_size, device):
+        return torch.multinomial(self.prob, batch_size, replacement=True).to(device)
+    

vvembed/scripts/convert_nnscaler_checkpoint_to_transformers.py

@@ -0,0 +1,166 @@
+#!/usr/bin/env python
+# coding=utf-8
+
+import argparse
+import json
+import os
+from pathlib import Path
+import re
+import torch
+from typing import Dict, List, Tuple
+
+from modular.configuration_vibevoice import (
+    VibeVoiceConfig
+)
+from modular.modeling_vibevoice import VibeVoiceForConditionalGeneration
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+def convert_vibevoice_nnscaler_checkpoint_to_hf(
+    checkpoint_path: str,
+    pytorch_dump_folder_path: str,
+    config_path: str = None,
+):
+    """
+    Convert a nnscaler VibeVoice checkpoint to HuggingFace format.
+    Supports both regular checkpoints and tensor parallel checkpoints.
+    """
+    
+    # Load regular checkpoint
+    logger.info(f"Loading regular checkpoint from {checkpoint_path}")
+    checkpoint = torch.load(checkpoint_path, map_location="cpu") # ['model', 'optimizer', 'lr_scheduler', 'train_status', 'train_args', 'rng_states', 'nnscaler', 'dataloader']
+    
+    # config = checkpoint['train_args']
+    init_config_name = checkpoint['train_args']['vars']['model_args']['config_path']['relative_path']
+    pretrained_name = checkpoint['train_args']['vars']['data_args']['tokenizer_path']
+    
+    init_config_path = Path(__file__).parent.parent / 'configs' / init_config_name.split('/')[-1]
+    if init_config_path.exists():
+        logger.info(f"Loading initial config from {init_config_path}")
+        with open(init_config_path, 'r') as f:
+            init_config = json.load(f)
+    else:
+        raise FileNotFoundError(f"Initial config file {init_config_path} not found. Please provide a valid path.")
+
+    tie_word_embeddings = init_config['decoder_config'].get('tie_word_embeddings', True)
+    logger.info(f"Tie word embeddings: {tie_word_embeddings}")
+
+    init_config['decoder_config']['use_cache'] = True
+    config = VibeVoiceConfig(**init_config, tie_word_embeddings=tie_word_embeddings)
+
+    # # Extract the model state dict
+    model_state_dict = {k.replace('model.model.', 'model.'): v for k, v in checkpoint["model"].items() if k.startswith('model.model.')}
+    if not tie_word_embeddings and 'model.lm_head.weight' in checkpoint["model"].keys():
+        # If not tying weights, we need to add the lm_head weight separately
+        model_state_dict['lm_head.weight'] = checkpoint["model"]['model.lm_head.weight']
+    
+    # Override with provided config if available
+    if config_path:
+        logger.info(f"Loading config from {config_path}")
+        with open(config_path, 'r') as f:
+            config_dict = json.load(f)
+        config = VibeVoiceConfig.from_dict(config_dict)
+    
+    # Set the default dtype to bfloat16 before creating the model
+    original_dtype = torch.get_default_dtype()
+    torch.set_default_dtype(torch.bfloat16)
+
+    # Create the HuggingFace model
+    logger.info("Creating HuggingFace VibeVoiceForConditionalGeneration model")
+    model = VibeVoiceForConditionalGeneration(config)
+    
+    # Restore original dtype
+    torch.set_default_dtype(original_dtype)
+
+    # Load the state dict
+    logger.info("Loading weights into model")
+    missing_keys, unexpected_keys = model.load_state_dict(model_state_dict, strict=False)
+    
+    if missing_keys:
+        logger.warning(f"Missing keys: {missing_keys}")
+    if unexpected_keys:
+        logger.warning(f"Unexpected keys: {unexpected_keys}")
+    
+    # Create output directory
+    os.makedirs(pytorch_dump_folder_path, exist_ok=True)
+    
+    # Save the model and config
+    logger.info(f"Saving model to {pytorch_dump_folder_path}")
+    
+    # Save config
+    config.save_pretrained(pytorch_dump_folder_path)
+    
+    # Save VibeVoiceProcessor configuration
+    logger.info("Saving VibeVoiceProcessor configuration")
+    processor_config = {
+        "processor_class": "VibeVoiceProcessor",
+        "speech_tok_compress_ratio": 3200,
+        "db_normalize": True,
+        # Audio processor configuration
+        "audio_processor": {
+            "feature_extractor_type": "VibeVoiceTokenizerProcessor",
+            "sampling_rate": 24000,
+            "normalize_audio": True,
+            "target_dB_FS": -25,
+            "eps": 1e-6,
+        },
+        "language_model_pretrained_name": pretrained_name,
+    }
+    
+    processor_config_path = os.path.join(pytorch_dump_folder_path, "preprocessor_config.json")
+    with open(processor_config_path, 'w') as f:
+        json.dump(processor_config, f, indent=2)
+    logger.info(f"Saved processor config to {processor_config_path}")
+    
+    # Save model with sharding
+    # save_pretrained handles tied weights automatically
+    logger.info("Saving model weights with sharding...")
+    model.save_pretrained(
+        pytorch_dump_folder_path,
+        max_shard_size="2GB",  # Set maximum size for each shard
+        safe_serialization=True  # Ensure saving in .safetensors format
+    )
+    logger.info(f"Model weights saved to {pytorch_dump_folder_path}")
+    
+    logger.info("Conversion complete!")
+    
+    # Verify the saved model can be loaded
+    logger.info("Verifying saved model...")
+    loaded_model = VibeVoiceForConditionalGeneration.from_pretrained(pytorch_dump_folder_path)
+    logger.info("Model successfully loaded from saved checkpoint!")
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--nnscaler_checkpoint_path",
+        type=str,
+        required=True,
+        help="Path to the fairseq checkpoint (.pt file). For tensor parallel checkpoints, "
+             "provide any one of the part files (e.g., checkpoint_1_5000-model_part-0.pt), "
+             "and the script will automatically detect and merge all parts.",
+    )
+    parser.add_argument(
+        "--pytorch_dump_folder_path", 
+        type=str,
+        required=True,
+        help="Path to the output PyTorch model directory",
+    )
+    parser.add_argument(
+        "--config_path",
+        type=str,
+        default=None,
+        help="Optional path to a config JSON file to override extracted config",
+    )
+    
+    args = parser.parse_args()
+    
+    convert_vibevoice_nnscaler_checkpoint_to_hf(
+        args.nnscaler_checkpoint_path,
+        args.pytorch_dump_folder_path,
+        args.config_path,
+    )
+
+
+if __name__ == "__main__":
+    main()