tts.py

import torch
from transformers import SpeechT5ForTextToSpeech, SpeechT5Processor
import logging
import numpy as np
import soundfile as sf
import torchaudio

# Set up logging
logging.basicConfig(level=logging.DEBUG)

MODEL_ID = "microsoft/speecht5_tts"

# Try to load the model and processor
try:
    processor = SpeechT5Processor.from_pretrained(MODEL_ID)
    model = SpeechT5ForTextToSpeech.from_pretrained(MODEL_ID)
    logging.info("Model and processor loaded successfully.")
except Exception as e:
    logging.error(f"Error loading model or processor: {e}")

def synthesize_speech(text):
    try:
        inputs = processor(text, return_tensors="pt")
        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        model.to(device)
        inputs = inputs.to(device)

        with torch.no_grad():
            speech = model.generate(**inputs)

        logging.info("Speech generated successfully.")

        # Decode the generated speech and save to an audio file
        waveform = speech.cpu().numpy().flatten()
        # Use torchaudio to save the waveform
        file_path = "output.wav"
        torchaudio.save(file_path, torch.tensor(waveform).unsqueeze(0), 16000)
        logging.info(f"Audio file saved successfully at {file_path}.")
        return file_path
    except Exception as e:
        logging.error(f"Error during speech synthesis: {e}")
        return None