app.py

import gradio as gr
from supertonic import TTS
import os

# Initialize TTS - auto_download=True handles the HF model fetching automatically
tts = TTS(auto_download=True)

# List of available voices based on the repository structure
VOICES = ["M1", "M2", "M3", "M4", "M5", "F1", "F2", "F3", "F4", "F5"]

# List of supported languages (Top 10 shown, you can add all 31)
LANGUAGES = {
    "English": "en", "Korean": "ko", "Japanese": "ja", "Spanish": "es",
    "French": "fr", "German": "de", "Hindi": "hi", "Italian": "it", 
    "Portuguese": "pt", "Russian": "ru", "Vietnamese": "vi"
}

def generate_speech(text, voice, language_name):
    lang_code = LANGUAGES[language_name]
    
    # Get the voice style object
    style = tts.get_voice_style(voice_name=voice)
    
    # Synthesize
    wav, duration = tts.synthesize(text, voice_style=style, lang=lang_code)
    
    # Save to a temporary path for Gradio to pick up
    output_path = "output.wav"
    tts.save_audio(wav, output_path)
    
    return output_path, f"Duration: {duration:.2f} seconds"

# Define the Gradio Interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("# 🎙️ Supertonic 3 TTS (On-Device CPU)")
    gr.Markdown("Lightning-fast multilingual TTS by Supertone. Running entirely on CPU via ONNX.")
    
    with gr.Row():
        with gr.Column():
            input_text = gr.Textbox(
                label="Input Text", 
                placeholder="Type something here...",
                value="Hello! Supertonic 3 is now running on a Hugging Face CPU Space.",
                lines=3
            )
            voice_opt = gr.Dropdown(choices=VOICES, value="M1", label="Voice Style")
            lang_opt = gr.Dropdown(choices=list(LANGUAGES.keys()), value="English", label="Language")
            btn = gr.Button("Generate Audio", variant="primary")
            
        with gr.Column():
            audio_output = gr.Audio(label="Synthesized Audio", type="filepath")
            stats = gr.Label(label="Metadata")

    gr.Examples(
        examples=[
            ["The train delay was announced at 4:45 PM <breath> due to track maintenance.", "M1", "English"],
            ["こんにちは、スーパートーンの世界へようこそ。", "F1", "Japanese"],
            ["¡Hola! Este es un ejemplo de síntesis de voz en español.", "M2", "Spanish"]
        ],
        inputs=[input_text, voice_opt, lang_opt]
    )

    btn.click(generate_speech, inputs=[input_text, voice_opt, lang_opt], outputs=[audio_output, stats])

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