app.py

import sys
from pathlib import Path

APP_ROOT = Path(__file__).resolve().parent
if str(APP_ROOT) not in sys.path:
    sys.path.insert(0, str(APP_ROOT))

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

try:
    from TTS.tts.utils.monotonic_align import core as _monotonic_align_core  # noqa: F401
except ImportError as exc:
    raise RuntimeError(
        "monotonic_align extension is missing. Run `pip install -e .` during build to compile XTTS dependencies."
    ) from exc

# Get device
device = "cuda" if torch.cuda.is_available() else "cpu"

# Init TTS
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)

def voice_clone(text: str, speaker_wav: str, language: str):
    # Run TTS
    print("Speaker wav:", speaker_wav)
    tts.tts_to_file(text=text, speaker_wav=speaker_wav, language=language, file_path="output.wav")
    return "output.wav"

iface = gr.Interface(fn=voice_clone, theme="Nymbo/Nymbo_Theme", 
                     inputs=[gr.Textbox(lines=2, placeholder="Enter the text...", label="Text"),
                             gr.Audio(type="filepath", label="Upload audio file"),
                             gr.Radio(['ru', 'en', 'zh-cn', 'ja', 'de', 'fr', 'it', 'pt', 'pl', 'tr', 'ko', 'nl', 'cs', 'ar', 'es', 'hu'], label="language"),
                            ],
                     outputs=gr.Audio(type="filepath", label="Generated audio file"), 
                     title="Voice Cloning")

iface.launch()