coqui_tts.py

import os
import noisereduce as nr
import soundfile as sf
# from moviepy.editor import *
import string
import json
from glob import glob
import torchaudio
import subprocess
import shutil
import pyloudnorm as pyln
import torch
from TTS.api import TTS


def run_audio_generation_v1(text,accent='None'):
    gpu = True if torch.cuda.is_available() else False
    tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=gpu) # gpu should be true when server (cuda)

    # pre-process story audio file
    # convert to 16 bit mono
    # remove noise
    speaker_wav_data, speaker_wav_rate = sf.read("./tmp/audio/input_src/0.wav")
    speaker_wav_data_no_noise = nr.reduce_noise(y=speaker_wav_data, sr=speaker_wav_rate)
    sf.write('./tmp/audio/speaker_wav.wav', speaker_wav_data_no_noise, speaker_wav_rate, subtype='PCM_16')

    tts.tts_to_file(
        text,
        speaker_wav="./tmp/audio/speaker_wav.wav",
        language="en",
        file_path="audio/output.wav"
    )