quantumphysicistsam/SpeechGeneration

1import os2import tempfile3from fastapi import FastAPI, UploadFile, File, Form4from fastapi.responses import FileResponse, JSONResponse5from TTS.api import TTS6import uvicorn7 8# Initialize FastAPI9app = FastAPI(10    title="TTS Voice Cloning API",11    description="API endpoint for synthesizing speech using Coqui TTS's voice cloning model.",12    version="1.0"13)14 15# Load the voice cloning model (YourTTS) for supported languages.16# Supported languages: "en", "fr-fr", "pt-br"17model_your_tts = TTS(18    model_name="tts_models/multilingual/multi-dataset/your_tts",19    progress_bar=False,20    gpu=False  # Change to True if your deployment environment has a GPU21)22 23@app.post("/synthesize/", summary="Synthesize speech from text")24async def synthesize_speech(25    text: str = Form(..., description="The text to synthesize."),26    language: str = Form(..., description="Language code (e.g., en, fr-fr, pt-br)."),27    speaker_file: UploadFile = File(None, description="Optional speaker reference WAV file")28):29    """30    Synthesize speech from text using the TTS model.  31    If no speaker file is uploaded, the default speaker (`my_voice.wav`) will be used if available.32    """33    # Create a temporary file to store the generated audio.34    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:35        output_path = tmp_file.name36 37    # Process the uploaded speaker file if provided.38    if speaker_file:39        try:40            # Save the uploaded file to a temporary location.41            with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as spk_file:42                content = await speaker_file.read()43                spk_file.write(content)44                speaker_path = spk_file.name45        except Exception as e:46            return JSONResponse(status_code=400, content={"error": f"Error processing speaker file: {str(e)}"})47    else:48        # Use the default recorded voice if available.49        default_speaker = "my_voice.wav"50        speaker_path = default_speaker if os.path.exists(default_speaker) else None51 52    # Attempt to synthesize speech.53    try:54        model_your_tts.tts_to_file(55            text=text,56            language=language,57            speaker_wav=speaker_path,58            file_path=output_path59        )60    except Exception as e:61        return JSONResponse(status_code=500, content={"error": f"Error during synthesis: {str(e)}"})62 63    # Return the synthesized audio as a WAV file.64    return FileResponse(65        output_path,66        media_type="audio/wav",67        filename="output.wav"68    )69 70# Optional: Run the server locally (useful for testing)71if __name__ == "__main__":72    uvicorn.run(app, host="0.0.0.0", port=8000)73