TTS dialog generator

import torch
from TTS.api import TTS
from datetime import datetime
import os
from pydub import AudioSegment
import re


def generate_tts(output_folder, text, speaker, index):
    device = "cuda" if torch.cuda.is_available() else "cpu"
    tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
    if not os.path.exists(output_folder):
        os.mkdir(output_folder)
    filename = re.sub(r"[^A-Za-z0-9_\-\.]", "", text[:50].replace(' ', '_').replace('.', '').lower())
    output = f"{index}_{speaker}_{filename}.wav"
    if not os.path.exists(f'{output_folder}/{output}'):
        tts.tts_to_file(text=text, speaker_wav=f"voices/{speaker}.wav", language="fr", file_path=f'{output_folder}/{output}')

def merge_audio_files(output_folder):
    combined = AudioSegment.empty()
    folder = os.listdir(output_folder)
    for f in folder:
        if f.startswith('combined'):
            continue
        if f.endswith('.wav'):
            combined += AudioSegment.from_wav(f'{output_folder}/{f}')
    filename = f"{output_folder}/combined_{datetime.now().strftime('%Y%m%d_%H%M')}.wav"
    combined.export(filename, format="wav")


output_folder = 'outputs/example_dialog1'

dialog = [
    {"character": "franck", "text": "Bonjour, comment allez-vous?"},
    {"character": "female", "text": "Très bien, merci. Et vous?"},
]

i = 0
for line in dialog:
    speaker = line["character"]
    text = line["text"]
    generate_tts(output_folder, text, speaker, i)
    i+=1

merge_audio_files(output_folder)