app.py

import os
import torch
import gradio as gr
from openvoice import se_extractor
from openvoice.api import BaseSpeakerTTS, ToneColorConverter
import multiprocessing

# Function to initialize CUDA and run the main task in a subprocess
def initialize_and_run(input_text: str, save_path: str):
    try:
        ckpt_base = 'checkpoints/base_speakers/ZH'
        ckpt_converter = 'checkpoints/converter'
        device = "cuda:0"
        print(device)
        output_dir = 'outputs'

        base_speaker_tts = BaseSpeakerTTS(f'{ckpt_base}/config.json', device=device)
        base_speaker_tts.load_ckpt(f'{ckpt_base}/checkpoint.pth')

        tone_color_converter = ToneColorConverter(f'{ckpt_converter}/config.json', device=device)
        tone_color_converter.load_ckpt(f'{ckpt_converter}/checkpoint.pth')
        source_se = torch.load(f'{ckpt_base}/zh_default_se.pth').to(device)
        os.makedirs(output_dir, exist_ok=True)
        reference_speaker = './resources/demo_speaker0.mp3'  # This is the voice you want to clone
        target_se, audio_name = se_extractor.get_se(reference_speaker, tone_color_converter, target_dir='processed', vad=True)

        # Run the base speaker tts
        src_path = f'{output_dir}/tmp.wav'
        base_speaker_tts.tts(input_text, src_path, speaker='default', language='Chinese', speed=1.0)

        # Run the tone color converter
        encode_message = "@MyShell"
        tone_color_converter.convert(
            audio_src_path=src_path,
            src_se=source_se,
            tgt_se=target_se,
            output_path=save_path,
            message=encode_message)
    except Exception as e:
        print(f"Error during GPU task execution: {e}")
        raise

def audio_io(input_text: str):
    output_dir = 'outputs'
    save_path = f'{output_dir}/output_chinese.wav'
    process = multiprocessing.Process(target=initialize_and_run, args=(input_text, save_path))
    process.start()
    process.join()
    if process.exitcode != 0:
        raise RuntimeError("Subprocess for GPU task failed.")
    return save_path

demo = gr.Interface(
    fn=audio_io,
    inputs=["text"],
    outputs=["audio"],
)

if __name__ == "__main__":
    if torch.cuda.is_available():
        generator = torch.Generator('cuda').manual_seed(seed)
    else:
        generator = torch.Generator().manual_seed(seed)
    demo.launch(share=True)