tts_api.py

from flask import jsonify, send_file, request
from main import *
import torch, torchaudio, uuid, io, base64

def text_to_speech_func(text, output_path="output_audio.wav"):
    if tts_model is None: return {"error": "TTS model not initialized."}
    input_tokens = tts_model.tokenizer(text, return_tensors="pt", padding=True).to(device)
    with torch.no_grad(): audio_output = tts_model(input_tokens['input_ids'])
    torchaudio.save(output_path, audio_output.cpu(), 16000); return output_path

def tts_api(text):
    output_file = text_to_speech_func(text)
    if isinstance(output_file, dict) and "error" in output_file: return {"error": output_file["error"]}
    with open(output_file, 'rb') as f: audio_content = f.read()
    audio_base64 = base64.b64encode(audio_content).decode('utf-8'); os.remove(output_file); return {"audio_base64": audio_base64, "mimetype": "audio/wav"}