app.py

import os
import shutil
import torch
from TTS.api import TTS
import gradio as gr
from faster_whisper import WhisperModel

# Optional: run local setup.py if needed
import subprocess
try:
    subprocess.run(['python', 'setup.py', 'install', '--user'], check=True)
    print("Installation successful.")
except subprocess.CalledProcessError as e:
    print(f"Installation failed with error: {e}")

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

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

# Initialize Whisper model for transcription
whisper_model = WhisperModel("small", device=device)

# Folder for dataset
dataset_folder = "my_voice_dataset/audio"
os.makedirs(dataset_folder, exist_ok=True)

# Transcription function
def transcribe_language(audio_path: str) -> str:
    segments, _ = whisper_model.transcribe(audio_path)
    transcription = " ".join([seg.text for seg in segments])
    return transcription

# Voice cloning function
def voice_clone(text: str, speaker_wav: str):
    # Save uploaded audio
    filename = os.path.basename(speaker_wav)
    saved_path = os.path.join(dataset_folder, filename)
    shutil.copy(speaker_wav, saved_path)
    print(f"Saved uploaded audio to: {saved_path}")

    # Transcribe audio
    transcription = transcribe_language(saved_path)
    print(f"Transcription: {transcription}")

    # Detect language automatically (fallback to 'en')
    language = "en"
    if transcription.strip():
        try:
            from langdetect import detect
            language = detect(transcription)
            print(f"Detected language: {language}")
        except Exception as e:
            print(f"Language detection failed: {e}")

    # Generate speech
    tts.tts_to_file(text=text, speaker_wav=saved_path, language=language, file_path="output.wav")
    return "output.wav"

# Gradio interface
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"),
    ],
    outputs=gr.Audio(type="filepath", label="Generated audio file"),
    title="Voice Cloning with Automatic Language Detection"
)

iface.launch()