app.py

import os
os.environ["NUMBA_DISABLE_CACHE"] = "1"
import gradio as gr
from openvoice import OpenVoice

# Load OpenVoice model (adjust to your setup)
model = OpenVoice(language="en")

def clone_and_speak(audio, text):
    output_path = "output.wav"
    model.clone_voice(
        source_audio_path=audio.name,
        target_text=text,
        output_path=output_path
    )
    return output_path

with gr.Blocks() as demo:
    gr.Markdown("# OpenVoice TTS - Hugging Face Space")
    with gr.Row():
        audio_input = gr.Audio(label="Upload voice to clone", type="file")
        text_input = gr.Textbox(label="Enter text to synthesize")
    with gr.Row():
        generate_btn = gr.Button("Generate Audio")
        audio_output = gr.Audio(label="Synthesized Output", type="filepath")

    generate_btn.click(fn=clone_and_speak, inputs=[audio_input, text_input], outputs=audio_output)

demo.launch()