medahmedkrichen/ViDubb ? reverse-engineered prompt

Build me a simple web app for AI video dubbing that lets me upload a video or paste a YouTube link, then automatically transcribe it, translate it into another language, and generate a new dubbed version with a realistic cloned voice.

I want it to handle more than one speaker, keep the original background sounds if possible, and make the new audio match the mouth movements as closely as it can. It should also let me choose whether I want the result with the original background audio or without it. A clean Gradio style interface would be great, with a way to preview the input and download the finished video.

If you need to look up current docs or models online to make the lip sync and speech parts work better, go ahead. Please make it work in a notebook or as a small app I can run locally with a GPU if available, but still have a CPU fallback if not.