app.py

import os
import gradio as gr
import numpy as np
import soundfile as sf
import tempfile

# ------------------------------------
# REPO AUTO-CLONE SECTION (as you asked)
# ------------------------------------
REPO_URL = "https://huggingface.co/Supertone/supertonic"
TARGET_DIR = "supertonic"  # folder name after clone

def run_cmd(cmd):
    print(f"[CMD] {cmd}")
    return os.system(cmd)

print("=== Checking Supertonic repo ===")
if not os.path.exists(TARGET_DIR):
    print("[+] Cloning repo (LFS pointers only)...")
    run_cmd("git lfs install")
    run_cmd(f"GIT_LFS_SKIP_SMUDGE=1 git clone {REPO_URL} {TARGET_DIR}")
else:
    print("[✓] Repo already exists. Skipping clone.")


# ------------------------------------
# IMPORT TTS MODULE FROM YOUR CLONED PATH
# ------------------------------------
# You must make sure your tts_model.py is inside the repo or same folder
from tts_model import (
    load_text_to_speech,
    load_voice_style,
    sanitize_filename,
)

# ------------------------------------
# LOAD MODEL
# ------------------------------------
ONNX_DIR = "./onnx"              # change if needed
STYLE_FILES = ["./styles/default.json"]
TOTAL_STEP = 30

print("Loading TTS model...")
tts = load_text_to_speech(ONNX_DIR)
style = load_voice_style(STYLE_FILES)


# ------------------------------------
# TTS FUNCTION
# ------------------------------------
def run_tts(text, speed):
    if not text.strip():
        return None, "Text cannot be empty."

    wav, dur = tts(
        text=text,
        style=style,
        total_step=TOTAL_STEP,
        speed=float(speed),
        silence_duration=0.3,
    )

    wav = wav.squeeze()  # [1, N] -> [N]

    # Save to temporary WAV
    tmp_path = tempfile.mktemp(suffix=".wav")
    sf.write(tmp_path, wav, tts.sample_rate)

    return tmp_path, f"Generated duration: {float(dur):.2f}s"


# ------------------------------------
# GRADIO UI
# ------------------------------------
def ui():
    with gr.Blocks(title="ONNX TTS") as demo:
        gr.Markdown("## 🎤 ONNX Text-To-Speech Demo")

        text = gr.Textbox(
            label="Input Text",
            placeholder="Type something...",
            lines=4
        )

        speed = gr.Slider(
            label="Speed",
            minimum=0.5,
            maximum=2.0,
            value=1.05,
            step=0.01
        )

        generate_btn = gr.Button("Generate Speech 🔊")

        audio_out = gr.Audio(label="Output")
        info = gr.Textbox(label="Info", interactive=False)

        generate_btn.click(
            fn=run_tts,
            inputs=[text, speed],
            outputs=[audio_out, info]
        )

    return demo


app = ui()

if __name__ == "__main__":
    app.launch()