app.py

import gradio as gr
import uuid
import os
from datetime import timedelta
from TTS.api import TTS
import locale
import torch

locale.getpreferredencoding = lambda: "UTF-8"
device = "cuda" if torch.cuda.is_available() else "cpu"

os.environ["COQUI_TOS_AGREED"] = "1"
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
tts = TTS("xtts_v2.0.2")

SECRET_USER = "admin_kr"
SECRET_PASSWORD = "9NLa2)597,J^"


async def generate_audio(text_input: str, creator: str, top_k_input: int) -> str:

    refer_voices = ''
    unique_id = str(uuid.uuid4())

    output_file = f'output.wav'
    match creator:
        case 'Roomie':
            refer_voices = ["./assets/roomie/roomie_emocionado_base_1.wav", "./assets/roomie/ref_12.wav", "./assets/roomie/ref_11.wav", "./assets/roomie/ref_10.wav",
                            "./assets/roomie/ref_1.wav", "./assets/roomie/ref_6.wav", "./assets/roomie/ref_7.wav", "./assets/roomie/ref_8.wav", "./assets/roomie/roomie_emocionado_base_2.wav",]
        case 'Xavy':
            refer_voices = ["./assets/xavy/neutro_3.wav",
                            "./assets/xavy/neutro_1.wav", "./assets/xavy/neutro_2.wav"]
        case 'Bella':
            refer_voices = ["./assets/bella/neutro_2.wav",
                            "./assets/bella/neutro_1.wav", "./assets/bella/neutro_3.wav"]

        case 'Julia':
            refer_voices = ["assets/julia/neutro_4_Final_fast.wav", "assets/julia/enfadado_1_Final.wav", "assets/julia/enfadado_2_Final.wav",
                            "assets/julia/enfadado_3_Final.wav", "assets/julia/emocionada_1.wav", "assets/julia/emocionada_2_Final.wav"]

    tts.tts_to_file(text=text_input,
                    file_path=output_file,
                    speaker_wav=refer_voices,
                    language="en",
                    split_sentences=True,
                    top_k=top_k_input
                    )

    source_audio_file_name = output_file

    return gr.Audio(value=source_audio_file_name)


app = gr.Interface(
    fn=generate_audio,
    inputs=[gr.Textbox(label='Text to Speach'), gr.Dropdown(
        ['Roomie', 'Xavy', 'Bella', 'Julia'], label="Coice your creator"),
        gr.Slider(0, 100, value=50, label='Emotion', info='Valores mas altos para tonos mas emocionado, valores bajos para tonos mas aburrido ')],

    outputs=['audio']
)

app.launch(auth=(SECRET_USER, SECRET_PASSWORD))