app.py

import gradio as gr
import torch
from TTS.api import TTS
import os

# Aceptar términos de uso de Coqui TTS
os.environ["COQUI_TOS_AGREED"] = "1"

# Configurar para usar CPU si no hay GPU disponible
device = "cuda" if torch.cuda.is_available() else "cpu"

# Inicializar el modelo de TTS con manejo seguro de carga
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=torch.cuda.is_available()).to(device)

# Función para clonar la voz y generar el archivo de audio
def clone(text, audio):
    output_path = "./output.wav"
    tts.tts_to_file(text=text, speaker_wav=audio, language="es", file_path=output_path)
    return output_path

# Interfaz de Gradio
iface = gr.Interface(
    fn=clone, 
    inputs=[
        gr.Textbox(label='Text'), 
        gr.Audio(type='filepath', label='Voice reference audio file')
    ], 
    outputs=gr.Audio(type='filepath'),
    title='cn-speech-esss',
    description="""
    by [Gitgato](gitgato)
    This space uses the xtts_v2 model. Non-commercial use only. [Coqui Public Model License](https://coqui.ai/cpml)
    
    Please ❤️ this Space. <a href="mailto:a@om">Email me</a>.
    """,
    theme=gr.themes.Base(primary_hue="teal", secondary_hue="teal", neutral_hue="slate"),
    examples=[
        ["Hey! It's me Dorthy, from the Wizard of Oz. Type in whatever you'd like me to say.", "./audio/Wizard-of-Oz-Dorthy.wav"],
        ["It's me Vito Corleone, from the Godfather. Type in whatever you'd like me to say.", "./audio/Godfather.wav"],
        ["Hey, it's me Paris Hilton. Type in whatever you'd like me to say.", "./audio/Paris-Hilton.mp3"],
        ["Hey, it's me Megan Fox from Transformers. Type in whatever you'd like me to say.", "./audio/Megan-Fox.mp3"],
        ["Hey there, it's me Jeff Goldblum. Type in whatever you'd like me to say.", "./audio/Jeff-Goldblum.mp3"],
        ["Hey there, it's me Heath Ledger as the Joker. Type in whatever you'd like me to say.", "./audio/Heath-Ledger.mp3"],
    ]
)

# Lanzar la interfaz
iface.launch()