app.py

import gradio as gr
import sys, os
from huggingface_hub import snapshot_download, hf_hub_download

import torch
from cosyvoice.utils.file_utils import load_wav
from uuid import uuid1
import uuid
from cosyvoice_rodis.cli.cosyvoice import CosyVoice as CosyVoiceTTS_speakerminus
from pydub import AudioSegment
import tempfile
import soundfile as sf
import subprocess
import numpy as np
import random
import numpy


# import imageio_ffmpeg

# ffmpeg_path = imageio_ffmpeg.get_ffmpeg_exe()
# print(f"FFmpeg path: {ffmpeg_path}")
# user_bin = os.path.expanduser("~/bin")
# if not os.path.exists(user_bin):
#     os.makedirs(user_bin)
# ffmpeg_link = os.path.join(user_bin, "ffmpeg")
# if os.path.exists(ffmpeg_link):
#     os.remove(ffmpeg_link) 
# os.symlink(ffmpeg_path, ffmpeg_link)
# print(f"create symbolic link: {ffmpeg_link}")
# os.environ["PATH"] = f"{user_bin}:{os.environ.get('PATH', '')}"


sys.path.append('third_party/Matcha-TTS')
os.system('export PYTHONPATH=third_party/Matcha-TTS')

from huggingface_hub import hf_hub_download

# Download assets and logos in background to avoid blocking startup
assets_dir = None
logo_path = None
logo_path2 = None

def load_assets():
    """Load assets lazily"""
    global assets_dir, logo_path, logo_path2
    if assets_dir is None:
        try:
            print("Downloading assets and logos...")
            assets_dir = snapshot_download(
                repo_id="tienfeng/prompt",
                repo_type="dataset",
            )
            logo_path = hf_hub_download(
                repo_id="tienfeng/prompt",
                filename="logo2.png",
                repo_type="dataset",
            )
            logo_path2 = hf_hub_download(
                repo_id="tienfeng/prompt",
                filename="logo.png",
                repo_type="dataset",
            )
            print("Assets downloaded successfully")
        except Exception as e:
            print(f"Warning: Failed to download assets/logos: {e}")
            assets_dir = None
            logo_path = None
            logo_path2 = None

# Start downloading assets in background (non-blocking)
import threading
import time
assets_download_thread = threading.Thread(target=load_assets, daemon=True)
assets_download_thread.start()

# Wait for assets to download (with timeout) before creating UI
# This ensures logo is available when UI is created
max_wait_time = 30  # Maximum wait time in seconds
wait_interval = 0.5  # Check every 0.5 seconds
elapsed = 0
while logo_path is None and elapsed < max_wait_time:
    time.sleep(wait_interval)
    elapsed += wait_interval
if logo_path is None:
    print("Warning: Logo download timed out, UI will be created without logo")

# Delay model download to avoid blocking startup
model_repo_id = "AIDC-AI/Marco-Voice"
local_model = None
local_model_path = None
local_model_path_enhenced = None

def load_models():
    """Load models lazily when needed"""
    global local_model, local_model_path, local_model_path_enhenced
    if local_model is None:
        print("Downloading models...")
        local_model = snapshot_download(
            repo_id=model_repo_id,
            repo_type="model"
            # token=os.getenv("HF_TOKEN")
        )
        local_model_path = os.path.join(local_model, "marco_voice")
        local_model_path_enhenced = os.path.join(local_model, "marco_voice_enhenced")
        print("Models downloaded successfully")

# Delay model loading to avoid blocking startup
# Models will be loaded lazily when first used
tts_speakerminus = None
tts_sft = None

text_prompt = {
    "翟佳宁": "这个节目就是把四个男嘉宾，四个女嘉宾放一个大别墅里让他们朝夕相处一整个月，月末选择心动的彼此。",
    "范志毅": "没这个能力知道吗，我已经说了，你像这样的比赛本身就没有打好基础。",
    "呼兰": "发完之后那个工作人员说，老师，呼兰老师你还要再加个标签儿，我说加什么标签儿，他说你就加一个呼兰太好笑了。",
    "江梓浩": "就是很多我们这帮演员一整年也就上这么一个脱口秀类型的节目。",
    "李雪琴": "我就劝他，我说你呀，你没事儿也放松放松，你那身体都亮红灯儿了你还不注意。",
    "刘旸": "比如这三年我在街上开车，会在开车的时候进行一些哲思，我有天开车的时候路过一个地方。",
    "唐香玉": "大家好我叫唐香玉， 我年前把我的工作辞了，成了一个全职脱口秀演员。",
    "小鹿":  "然后我就老家的亲戚太多了，我也记不清谁该叫谁，所以我妈带着我和我。",
    "于祥宇": "我大学专业学的是哲学，然后节目组就说那这期主题你可以刚好聊一下哲学专业毕业之后的就业方向。",
    "赵晓卉": "终于没有人问我为什么不辞职了，结果谈到现在，谈恋爱第一天人家问我，能打个电话吗？我说你有啥事儿。",
    "徐志胜": "最舒服的一个方式，这个舞台也不一定就是说是来第一年就好嘛，只要你坚持，肯定会有发光发热的那天嘛。"
} 
audio_prompt = {
    "翟佳宁": "zhaijianing",
    "范志毅": "fanzhiyi",
    "呼兰": "hulan",
    "江梓浩": "jiangzhihao",
    "李雪琴": "lixueqin",
    "刘旸": "liuchang",
    "唐香玉": "tangxiangyu",
    "小鹿": "xiaolu",
    "于祥宇": "yuxiangyu",
    "赵晓卉": "zhaoxiaohui",
    "徐志胜": "xuzhisheng"
}
# audio_prompt_path = assets_dir

def load_audio_and_convert_to_16bit(file_path, target_sample_rate=16000):
    audio = AudioSegment.from_file(file_path)
    
    if audio.channels > 1:
        audio = audio.set_channels(1)
    
    if audio.frame_rate != target_sample_rate:
        audio = audio.set_frame_rate(target_sample_rate)
    
    audio_data = np.array(audio.get_array_of_samples(), dtype=np.float32)
    
    audio_data = audio_data / np.max(np.abs(audio_data))
    
    audio_data = (audio_data * 32767).astype(np.int16)
    
    return torch.tensor(audio_data), target_sample_rate

def convert_audio_with_sox(input_file, output_file, target_sample_rate=16000):
    try:
        # command = [
        #     './ffmpeg-7.0.2-amd64-static/ffmpeg', input_file,  
        #     '-r', str(target_sample_rate),  
        #     '-b', '16',  
        #     '-c', '1',
        #     output_file  
        # ]
        command = [
            './ffmpeg-7.0.2-amd64-static/ffmpeg',
            '-i', input_file,  
            '-ar', str(target_sample_rate), 
            '-ac', '1', 
            '-b:a', '16k', 
            '-f', 'wav', 
            output_file
        ]
        
        subprocess.run(command, check=True)
        print(f"Audio converted successfully: {output_file}")
    except subprocess.CalledProcessError as e:
        print(f"Error during conversion: {e}")

os.makedirs("./tmp", exist_ok=True)

def generate_speech_speakerminus(tts_text, speed, speaker, key, ref_audio, ref_text):
    # import pdb;pdb.set_trace()
    global tts_speakerminus_global, local_model_path
    # Ensure models are downloaded (this may take time on first use)
    if local_model_path is None:
        print("Downloading models (this may take a few minutes on first use)...")
        load_models()
    if 'tts_speakerminus_global' not in globals() or tts_speakerminus_global is None:
        print("Loading CosyVoice (speakerminus) model...")
        tts_speakerminus_global = CosyVoiceTTS_speakerminus(model_dir=local_model_path)

    if not ref_audio or not ref_text:
        # Ensure assets are loaded
        if assets_dir is None:
            load_assets()
        audio_prompt_path = assets_dir
        if audio_prompt_path is None or assets_dir is None:
            raise ValueError("Audio prompt path is not available. Please wait a moment and try again, or provide reference audio and text.")
        ref_text = text_prompt.get(speaker, "")
        speaker_audio_name = audio_prompt.get(speaker)
        if speaker_audio_name:
            ref_audio = os.path.join(audio_prompt_path, f"{speaker_audio_name}.wav")
        else:
            raise ValueError(f"Speaker '{speaker}' not found in audio_prompt dictionary")
    else:
        try:
            info = sf.info(ref_audio)
            sample_rate = info.samplerate
            channels = info.channels
            
            if sample_rate != 16000:
                raise ValueError(f"Invalid audio sample rate. Expected: 16000 Hz, got: {sample_rate} Hz. Please use a 16kHz audio file.")
            if channels != 1:
                raise ValueError(f"Invalid audio channel count. Expected: 1 (mono), got: {channels}. Please use a mono audio file.")
            
            file_ext = os.path.splitext(ref_audio)[1].lower()
            if file_ext != '.wav':
                raise ValueError(f"Invalid audio format. Expected: WAV format, got: {file_ext}. Please use a WAV format audio file.")
                
        except Exception as e:
            if isinstance(e, ValueError):
                raise e
            else:
                raise ValueError(f"Failed to read audio file: {str(e)}. Please ensure the audio file is in the correct format (16kHz mono WAV format).")


    if not ref_audio:
        raise ValueError("Reference audio is required but not provided")
    ref_audio = load_wav(ref_audio, 16000)
    emo = {"Sad": "伤心", "Fearful": "恐惧", "Happy": "快乐", "Surprise": "惊喜", "Angry": "生气", "Jolliest": "戏谑"} 
    # key="快乐"
    emotion_file = "./emotion_info.pt"
    if not os.path.exists(emotion_file):
        raise FileNotFoundError(f"Emotion info file not found: {emotion_file}. Please ensure this file exists in the workspace.")
    emotion_data = torch.load(emotion_file)
    if key in ["Angry", "Surprise", "Happy"]:
        emotion_info = emotion_data["male005"][key] 
    elif key in ["Sad"]:
        emotion_info = emotion_data["female005"][key]
    elif key in ["Fearful"]:
        emotion_info = emotion_data["female003"][key]
    else:
        emotion_info = emotion_data["male005"][key]

    sample_rate, full_audio = tts_speakerminus_global.synthesize(
        tts_text,
        prompt_text = ref_text,
        # speaker=speaker,
        prompt_speech_16k = ref_audio,
        key = emo.get(key),
        emotion_embedding=emotion_info,
        # ref_audio = ref_audio,
        # speed=speed
        
    )
    print("sample_rate:", sample_rate, "full_audio:", full_audio.min(), full_audio.max())
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio_file:
        output_audio_path = temp_audio_file.name
        audio_segment = AudioSegment(
            full_audio.tobytes(), 
            frame_rate=sample_rate,
            sample_width=full_audio.dtype.itemsize, 
            channels=1
        )
        audio_segment.export(output_audio_path, format="wav")
    print(f"Audio saved to {output_audio_path}")
    return output_audio_path


def generate_speech_sft(tts_text, speed, speaker, key, ref_audio, ref_text):
    global tts_sft_global, local_model_path_enhenced
    # Ensure models are downloaded (this may take time on first use)
    if local_model_path_enhenced is None:
        print("Downloading models (this may take a few minutes on first use)...")
        load_models()
    if 'tts_sft_global' not in globals() or tts_sft_global is None:
        print("Loading CosyVoice (SFT enhanced) model...")
        tts_sft_global = CosyVoiceTTS_speakerminus(model_dir=local_model_path_enhenced)
    if not ref_audio and not ref_text:
        # Ensure assets are loaded
        if assets_dir is None:
            load_assets()
        audio_prompt_path = assets_dir
        if audio_prompt_path is None or assets_dir is None:
            raise ValueError("Audio prompt path is not available. Please wait a moment and try again, or provide reference audio and text.")
        ref_text = text_prompt.get(speaker, "")
        speaker_audio_name = audio_prompt.get(speaker)
        if speaker_audio_name:
            ref_audio = os.path.join(audio_prompt_path, f"{speaker_audio_name}.wav")
        else:
            raise ValueError(f"Speaker '{speaker}' not found in audio_prompt dictionary")
    else:
        try:
            info = sf.info(ref_audio)
            sample_rate = info.samplerate
            channels = info.channels
            
            if sample_rate != 16000:
                raise ValueError(f"Invalid audio sample rate. Expected: 16000 Hz, got: {sample_rate} Hz. Please use a 16kHz audio file.")
            if channels != 1:
                raise ValueError(f"Invalid audio channel count. Expected: 1 (mono), got: {channels}. Please use a mono audio file.")
            
            file_ext = os.path.splitext(ref_audio)[1].lower()
            if file_ext != '.wav':
                raise ValueError(f"Invalid audio format. Expected: WAV format, got: {file_ext}. Please use a WAV format audio file.")
                
        except Exception as e:
            if isinstance(e, ValueError):
                raise e
            else:
                raise ValueError(f"Failed to read audio file: {str(e)}. Please ensure the audio file is in the correct format (16kHz mono WAV format).")
    if not ref_audio:
        raise ValueError("Reference audio is required but not provided")
    ref_audio = load_wav(ref_audio, 16000)

    emo = {"Sad": "伤心", "Fearful": "恐惧", "Happy": "快乐", "Surprise": "惊喜", "Angry": "生气", "Jolliest": "戏谑"} 
    # key="快乐"
    emotion_file = "./emotion_info.pt"
    if not os.path.exists(emotion_file):
        raise FileNotFoundError(f"Emotion info file not found: {emotion_file}. Please ensure this file exists in the workspace.")
    emotion_data = torch.load(emotion_file)
    if key in ["Angry", "Surprise", "Happy"]:
        emotion_info = emotion_data["male005"][key] 
    elif key in ["Sad"]:
        emotion_info = emotion_data["female005"][key]
    elif key in ["Fearful"]:
        emotion_info = emotion_data["female003"][key]
    else:
        emotion_info = emotion_data["male005"][key]

    sample_rate, full_audio = tts_sft_global.synthesize(
        tts_text,
        prompt_text = ref_text,
        # speaker=speaker,
        prompt_speech_16k = ref_audio,
        key = emo.get(key),
        emotion_embedding=emotion_info,
        # ref_audio = ref_audio,
        # speed=speed
        
    )
    print("sample_rate:", sample_rate, "full_audio:", full_audio.min(), full_audio.max())
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio_file:
        output_audio_path = temp_audio_file.name
        audio_segment = AudioSegment(
            full_audio.tobytes(), 
            frame_rate=sample_rate,
            sample_width=full_audio.dtype.itemsize, 
            channels=1
        )
        audio_segment.export(output_audio_path, format="wav")
    print(f"Audio saved to {output_audio_path}")
    return output_audio_path

names = [
    "于祥宇",
    "刘旸",
    "呼兰",
    "唐香玉",
    "小鹿",
    "李雪琴",
    "江梓浩",
    "翟佳宁",
    "范志毅",
    "赵晓卉",
    "徐志胜"
]

custom_css = """
:root {
    --primary-color: #6a11cb;
    --secondary-color: #2575fc;
    --accent-color: #ff6b6b;
    --light-bg: #f8f9fa;
    --dark-bg: #212529;
    --card-bg: #ffffff;
    --text-color: #343a40;
    --border-radius: 12px;
    --box-shadow: 0 6px 16px rgba(0,0,0,0.1);
    --transition: all 0.3s ease;
}

body {
    background: linear-gradient(135deg, var(--light-bg) 0%, #e9ecef 100%);
    min-height: 100vh;
    font-family: 'Segoe UI', 'PingFang SC', 'Microsoft YaHei', sans-serif;
    color: var(--text-color);
    line-height: 1.6;
}

.gradio-container {
    max-width: 1200px !important;
    margin: 2rem auto !important;
    padding: 0 1rem;
}

.header {
    padding: 0 !important;
    border-radius: 10px; /* 整个 Header 的圆角 */
    overflow: hidden;    /* 隐藏超出圆角部分的内容，非常重要！ */
}

/* 2. 行容器：使用 Flex 布局，让左右两部分等高 */
#header-row {
    display: flex;
    align-items: stretch; /* 让子项高度自动拉伸以填满容器 */
}

/* 3. Logo 容器：移除 Gradio 默认的所有样式，让它变成一个纯净的盒子 */
#logo-container {
    padding: 0 !important;
    border: none !important;
    background: none !important;
    box-shadow: none !important;
    min-width: 150px; /* 给 Logo 区域一个固定的最小宽度 */
    flex-shrink: 0;   /* 防止在空间不足时被压缩 */
}

/* 4. Logo 图片本身：让图片填满它的容器，就像背景图一样 */
#logo-container img {
    width: 100%;
    height: 100%;
    object-fit: cover; /* 裁剪并填充，保持宽高比，确保填满容器 */
    display: block;    /* 移除图片底部的微小空隙 */
}

/* 5. 标题区域：设置背景、内边距和文本对齐 */
#title-area {
    background: linear-gradient(to right, #5e57c2, #42a5f5); /* 右侧的渐变背景 */
    padding: 20px 25px; /* 给标题文本留出足够的空间 */
    display: flex;
    flex-direction: column;
    justify-content: center; /* 垂直居中标题 */
    flex-grow: 1; /* 占据所有剩余空间 */
}

/* 6. title */
#header-title h1 {
    color: white;
    font-size: 28px;
    margin: 0;
    font-weight: 600;
    text-shadow: 1px 1px 2px rgba(0,0,0,0.2);
}

.tabs {
    background: transparent !important;
    border: none !important;
    box-shadow: none !important;
}

.tab-nav {
    background: var(--card-bg) !important;
    border-radius: var(--border-radius) !important;
    padding: 0.5rem !important;
    margin-bottom: 1.5rem !important;
    box-shadow: var(--box-shadow) !important;
}

.tab-button {
    padding: 1rem 1.5rem !important;
    border-radius: 8px !important;
    font-weight: 600 !important;
    transition: var(--transition) !important;
    border: none !important;
}

.tab-button.selected {
    background: linear-gradient(135deg, var(--primary-color) 0%, var(--secondary-color) 100%) !important;
    color: white !important;
    box-shadow: 0 4px 12px rgba(106, 17, 203, 0.3) !important;
}

.tab-content {
    background: var(--card-bg) !important;
    border-radius: var(--border-radius) !important;
    padding: 2rem !important;
    box-shadow: var(--box-shadow) !important;
    margin-bottom: 2rem;
    border: none !important;
}

.input-section {
    background: #f9fafb;
    padding: 1.5rem;
    border-radius: var(--border-radius);
    margin-bottom: 1.5rem;
    border: 1px solid #e9ecef;
}

.output-section {
    background: #edf2f7;
    padding: 1.5rem;
    border-radius: var(--border-radius);
    border: 1px solid #e9ecef;
    display: flex;
    flex-direction: column;
    height: 100%;
}

.control-group {
    margin-bottom: 1.2rem;
}

.control-group label {
    display: block;
    margin-bottom: 0.5rem;
    font-weight: 600;
    color: #495057;
    font-size: 0.95rem;
}

input[type="text"], textarea {
    border-radius: 8px !important;
    padding: 0.8rem 1rem !important;
    border: 1px solid #ced4da !important;
    transition: var(--transition) !important;
}

input[type="text"]:focus, textarea:focus {
    border-color: var(--primary-color) !important;
    box-shadow: 0 0 0 3px rgba(106, 17, 203, 0.1) !important;
}

.slider {
    margin-top: 0.5rem !important;
}

.btn-generate {
    background: linear-gradient(135deg, var(--primary-color) 0%, var(--secondary-color) 100%) !important;
    color: white !important;
    font-weight: 600 !important;
    padding: 1rem 1.8rem !important;
    border-radius: 8px !important;
    border: none !important;
    transition: var(--transition) !important;
    font-size: 1rem !important;
    margin-top: auto;
    width: 100%;
    box-shadow: 0 4px 6px rgba(0,0,0,0.1) !important;
}

.btn-generate:hover {
    transform: translateY(-3px);
    box-shadow: 0 6px 12px rgba(106, 17, 203, 0.25) !important;
}

.example-text {
    background: #e9ecef;
    padding: 0.8rem;
    border-radius: 8px;
    font-style: italic;
    margin-top: 0.5rem;
    font-size: 0.9rem;
    color: #495057;
}

.audio-player {
    width: 100%;
    margin-top: 1rem;
    border-radius: 8px;
    overflow: hidden;
}

.model-info {
    background: #e6f7ff;
    padding: 1rem;
    border-radius: 8px;
    margin-top: 1.5rem;
    border-left: 4px solid #1890ff;
    font-size: 0.9rem;
}

.info-icon {
    color: #1890ff;
    margin-right: 8px;
    font-weight: bold;
}

.footer {
    text-align: center;
    color: #6c757d;
    font-size: 0.9rem;
    padding: 1.5rem 0;
    border-top: 1px solid #e9ecef;
    margin-top: 2rem;
}

.accordion {
    background: #f8f9fa !important;
    border-radius: 8px !important;
    padding: 0.8rem !important;
    margin-top: 1rem;
    border: 1px solid #e9ecef !important;
}

.accordion-title {
    font-weight: 600 !important;
    color: var(--primary-color) !important;
}

.audio-upload {
    border: 2px dashed #ced4da !important;
    border-radius: 8px !important;
    padding: 1.5rem !important;
    background: #f8f9fa !important;
    transition: var(--transition) !important;
}

.audio-upload:hover {
    border-color: var(--primary-color) !important;
    background: #f1f3f5 !important;
}

.audio-upload-label {
    font-weight: 500 !important;
    color: #495057 !important;
    margin-bottom: 0.5rem !important;
}

.radio-group {
    display: flex;
    flex-wrap: wrap;
    gap: 0.8rem;
    margin-top: 0.5rem;
}

.radio-item {
    flex: 1;
    min-width: 100px;
    text-align: center;
    padding: 0.8rem;
    border: 1px solid #ced4da;
    border-radius: 8px;
    cursor: pointer;
    transition: var(--transition);
}

.radio-item.selected {
    border-color: var(--primary-color);
    background: rgba(106, 17, 203, 0.05);
    color: var(--primary-color);
    font-weight: 500;
}

.radio-item:hover {
    border-color: var(--primary-color);
}

@media (max-width: 768px) {
    .header h1 {
        font-size: 2.2rem;
    }
    
    .header p {
        font-size: 1rem;
    }
    
    .gradio-container {
        padding: 0 0.5rem;
    }
    
    .tab-button {
        padding: 0.8rem 1rem !important;
        font-size: 0.9rem !important;
    }
}
"""


with gr.Blocks(css=custom_css, theme=gr.themes.Soft()) as demo:
    with gr.Column(elem_classes="header"):
        with gr.Row(elem_id="header-row", variant="compact"):
            # Load logo if available, otherwise hide the component to avoid showing upload interface
            # Note: gr.Image with value=None shows upload interface, so we use visible=False if no logo
            logo_exists = logo_path is not None and os.path.exists(logo_path) if logo_path else False
            logo_image = gr.Image(
                value=logo_path if logo_exists else None,
                elem_id="logo-container",
                show_label=False,
                show_download_button=False,
                show_share_button=False,
                visible=logo_exists
            ) 

            with gr.Column(elem_id="title-area"):
                gr.Markdown("# 🎤 Marco-Voice ", elem_id="header-title")
        
    with gr.Tabs(elem_classes="tabs") as tabs:
        with gr.TabItem("😄 Control of emotion", id=0):
            with gr.Row():
                with gr.Column(scale=2, elem_classes="input-section"):
                    gr.Markdown("### Input Settings")
                    tts_text_v1 = gr.Textbox(
                        lines=3, 
                        placeholder="Enter the text content you want to compose...",
                        label="Synthesizing text",
                        value="这真是太令人兴奋了！我们刚刚完成了一个重大突破！"
                    )
                    
                    with gr.Row():
                        with gr.Column():
                            speed_v1 = gr.Slider(
                                minimum=0.5, 
                                maximum=2.0, 
                                value=1.0, 
                                step=0.1, 
                                label="Speaking rate control"
                            )
                        with gr.Column():
                            emotion_v1 = gr.Radio(
                                choices=["Angry", "Happy", "Surprise", "Sad", "Fearful", "Jolliest"], 
                                value="Happy", 
                                label="Emotion selection"
                            )
                    
                    with gr.Row():
                        with gr.Column():
                            speaker_v1 = gr.Dropdown(
                                choices=names, 
                                value="徐志胜", 
                                label="Preset timbre"
                            )
                        with gr.Column():
                            gr.Markdown("### Or use a custom timbre")
                            with gr.Accordion("Upload reference audio", open=False, elem_classes="accordion"):
                                gr.Markdown("Upload 3-10 seconds of clear human voice as reference audio")
                                ref_audio_v1 = gr.Audio(
                                    type="filepath", 
                                    label="upload audio",
                                    elem_classes="audio-upload"
                                )
                                ref_text_v1 = gr.Textbox(
                                    lines=2, 
                                    placeholder="ref text content...",
                                    label="ref text"
                                )
                    
                    gr.Markdown("""
                    <div class="model-info">
                        <p><span class="info-icon">ℹ️</span> <strong>specification of a model：</strong> This model added emotion control ability on the basis of timbre cloning, and could generate speech with specific emotion.</p>
                        <p><span class="info-icon">💡</span> <strong>use skill：</strong> The sentiment expression effect is related to the content of the text, make sure the text matches the selected sentiment.</p>
                    </div>
                    """)
                
                with gr.Column(scale=1, elem_classes="output-section"):
                    gr.Markdown("### output result")
                    tts_v1_output = gr.Audio(
                        type="filepath", 
                        label="Generating speech",
                        interactive=False
                    )
                    tts_v1_button = gr.Button(
                        "🚀 Generating speech", 
                        variant="primary", 
                        elem_classes="btn-generate"
                    )
                    gr.Examples(
                        examples=[
                            ["这真是太令人兴奋了！我们刚刚完成了一个重大突破！", "Happy", "徐志胜"],
                            ["我简直不敢相信！这怎么可能发生？", "Surprise", "李雪琴"],
                            ["这太让人失望了，我们所有的努力都白费了。", "Sad", "范志毅"]
                        ],
                        inputs=[tts_text_v1, emotion_v1, speaker_v1],
                        label="Emotion example"
                    )
        with gr.TabItem("😄 Control of emotion enhenced", id=1):
            with gr.Row():
                with gr.Column(scale=2, elem_classes="input-section"):
                    gr.Markdown("### Input Settings")
                    tts_text_v2 = gr.Textbox(
                        lines=3, 
                        placeholder="Enter the text content you want to compose...",
                        label="Synthesizing text",
                        value="这真是太令人兴奋了！我们刚刚完成了一个重大突破！"
                    )
                    
                    with gr.Row():
                        with gr.Column():
                            speed_v2 = gr.Slider(
                                minimum=0.5, 
                                maximum=2.0, 
                                value=1.0, 
                                step=0.1, 
                                label="Speaking rate control"
                            )
                        with gr.Column():
                            emotion_v2 = gr.Radio(
                                choices=["Angry", "Happy", "Surprise", "Sad", "Fearful", "Jolliest"], 
                                value="Happy", 
                                label="Emotion selection"
                            )
                    
                    with gr.Row():
                        with gr.Column():
                            speaker_v2 = gr.Dropdown(
                                choices=names, 
                                value="徐志胜", 
                                label="Preset timbre"
                            )
                        with gr.Column():
                            gr.Markdown("### Or use a custom timbre")
                            with gr.Accordion("Upload reference audio", open=False, elem_classes="accordion"):
                                gr.Markdown("Upload 3-10 seconds of clear human voice as reference audio")
                                ref_audio_v2 = gr.Audio(
                                    type="filepath", 
                                    label="upload audio",
                                    elem_classes="audio-upload"
                                )
                                ref_text_v2 = gr.Textbox(
                                    lines=2, 
                                    placeholder="ref text content...",
                                    label="ref text"
                                )
                    
                    gr.Markdown("""
                    <div class="model-info">
                        <p><span class="info-icon">ℹ️</span> <strong>specification of a model：</strong> This model added emotion control ability on the basis of timbre cloning, and could generate speech with specific emotion.</p>
                        <p><span class="info-icon">💡</span> <strong>use skill：</strong> The sentiment expression effect is related to the content of the text, make sure the text matches the selected sentiment.</p>
                    </div>
                    """)
                
                with gr.Column(scale=1, elem_classes="output-section"):
                    gr.Markdown("### output result")
                    tts_v2_output = gr.Audio(
                        type="filepath", 
                        label="Generating speech",
                        interactive=False
                    )
                    tts_v2_button = gr.Button(
                        "🚀 Generating speech", 
                        variant="primary", 
                        elem_classes="btn-generate"
                    )
                    gr.Examples(
                        examples=[
                            ["这真是太令人兴奋了！我们刚刚完成了一个重大突破！", "Happy", "徐志胜"],
                            ["我简直不敢相信！这怎么可能发生？", "Surprise", "李雪琴"],
                            ["这太让人失望了，我们所有的努力都白费了。", "Sad", "范志毅"]
                        ],
                        inputs=[tts_text_v2, emotion_v2, speaker_v2],
                        label="emotion example"
                    )
    
    gr.Markdown("""
    <div class="footer">
        <p>Marco-Voice Text-to-Speech v1.0 | Based on an excellent open-source TTS model | Tech support:*****</p>
        <p>⚠️ Note: Synthesized speech is for technical demonstration and non-commercial use only.</p>
    </div>
    """)
    
    
    tts_v1_button.click(
        fn=generate_speech_speakerminus,
        inputs=[tts_text_v1, speed_v1, speaker_v1, emotion_v1, ref_audio_v1, ref_text_v1],
        outputs=tts_v1_output
    )
     # tts_text, speed, speaker, key, ref_audio, ref_text
    tts_v2_button.click(
        fn=generate_speech_sft,
        inputs=[tts_text_v2, speed_v2, speaker_v2, emotion_v2, ref_audio_v2, ref_text_v2],
        outputs=tts_v2_output
    )

# Don't preload models - let them download on first use to avoid startup timeout
# Models will be downloaded and loaded lazily when first requested by user  

if __name__ == "__main__":
    # Use environment variable for port (Hugging Face Spaces uses 7860 by default)
    server_port = int(os.environ.get("SERVER_PORT", 7860))
    launch_kwargs = {
        "server_name": "0.0.0.0",
        "server_port": server_port,
        "share": False,
    }
    # Only add favicon if it was successfully downloaded
    if logo_path is not None or logo_path2 is not None:
        launch_kwargs["favicon_path"] = logo_path
    demo.launch(**launch_kwargs)