import gradio as gr
import os

def generate_lipsync(image, audio):
    os.makedirs("input", exist_ok=True)
    os.makedirs("output", exist_ok=True)

    image_path = "input/image.jpg"
    audio_path = "input/audio.wav"
    video_path = "input/video.mp4"
    result_path = "output/result_voice.mp4"

    image.save(image_path)

    os.system(f"cp '{audio}' {audio_path}")

    os.system(
        f"ffmpeg -y -loop 1 -i {image_path} "
        f"-c:v libx264 -t 5 -pix_fmt yuv420p "
        f"-vf scale=512:512 {video_path}"
    )

    command = f"""
    python wav2lip/inference.py \
    --checkpoint_path checkpoints/wav2lip_gan.pth \
    --face {video_path} \
    --audio {audio_path} \
    --outfile {result_path}
    """

    os.system(command)

    return result_path


iface = gr.Interface(
    fn=generate_lipsync,
    inputs=[
        gr.Image(type="pil", label="Upload Image"),
        gr.Audio(type="filepath", label="Upload Audio")
    ],
    outputs=gr.Video(label="Generated Video"),
    title="TikTok AI LipSync"
)

iface.launch(server_name="0.0.0.0", server_port=7860, share=True)