| import os |
| import gradio as gr |
| import torch |
| import numpy as np |
|
|
| |
| os.environ["COQUI_TOS_AGREED"] = "1" |
|
|
| |
| original_torch_load = torch.load |
|
|
| def patched_torch_load(f, *args, **kwargs): |
| kwargs["weights_only"] = False |
| return original_torch_load(f, *args, **kwargs) |
|
|
| torch.load = patched_torch_load |
|
|
| |
| from TTS.api import TTS |
| tts = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2") |
|
|
| |
| def generate_audio(text, language, speaker_wav): |
| if speaker_wav is not None: |
| audio = tts.tts(text=text, speaker_wav=speaker_wav, language=language) |
| audio_np = np.array(audio, dtype=np.float16) |
| return (24000, audio_np) |
| |
|
|
| |
| iface = gr.Interface( |
| fn=generate_audio, |
| inputs=[ |
| gr.Text(label="Texto"), |
| gr.Text(label="Idioma (ej: 'es', 'en')"), |
| gr.Audio(type="filepath", label="Audio de voz") |
| ], |
| outputs="audio" |
| ) |
|
|
| iface.launch() |
|
|
|
|
