Spaces:
Sleeping
Sleeping
| from fastapi import FastAPI, Query, HTTPException | |
| from fastapi.responses import StreamingResponse | |
| from TTS.api import TTS | |
| import os | |
| from io import BytesIO | |
| from typing import Generator | |
| app = FastAPI() | |
| # Initialize the TTS model | |
| tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=False) # Set gpu=True if you have GPU support | |
| # Function to split text into chunks | |
| def split_text(text: str, words_per_chunk: int = 20): | |
| words = text.split() | |
| return [' '.join(words[i:i + words_per_chunk]) for i in range(0, len(words), words_per_chunk)] | |
| # Function to generate audio chunks | |
| def generate_audio_chunks( | |
| text: str, speaker_wav: str, language: str, chunk_size: int = 20 | |
| ) -> Generator[bytes, None, None]: | |
| if tts.is_multi_lingual and not language: | |
| raise ValueError("Language must be specified for multi-lingual models.") | |
| text_chunks = split_text(text, chunk_size) | |
| for idx, chunk in enumerate(text_chunks): | |
| # Generate audio for each chunk and yield as bytes | |
| audio_buffer = BytesIO() | |
| tts.tts_to_file( | |
| text=chunk, | |
| file_path=audio_buffer, | |
| speaker_wav=speaker_wav, | |
| language=language | |
| ) | |
| audio_buffer.seek(0) | |
| yield audio_buffer.read() | |
| async def generate_audio( | |
| text: str = Query(..., description="The input text to convert to speech."), | |
| language: str = Query("en", description="Language code for TTS (e.g., 'en' for English)."), | |
| speaker_wav: str = Query(..., description="Path to the WAV file for voice cloning.") | |
| ): | |
| if not os.path.exists(speaker_wav): | |
| raise HTTPException(status_code=400, detail="Speaker WAV file not found.") | |
| # StreamingResponse to stream audio chunks | |
| def audio_stream(): | |
| try: | |
| for audio_chunk in generate_audio_chunks(text=text, speaker_wav=speaker_wav, language=language): | |
| yield audio_chunk | |
| except Exception as e: | |
| raise HTTPException(status_code=500, detail=str(e)) | |
| return StreamingResponse(audio_stream(), media_type="audio/wav") | |