MoZoHuJa/ScarDeeX ? reverse-engineered prompt

Build me a simple local app for this LongCat video generation project so I can try it without messing around in the terminal.

I want one clean interface where I can choose between text to video, image to video, video continuation, long video generation, and the avatar audio driven modes if the model supports them. Make it easy to upload a prompt, an image, a starting video, or audio, then generate a playable result and let me download it. If there’s already a Streamlit demo or separate demo scripts, connect them into one straightforward experience and make the default settings sensible. Please also add clear setup instructions for installing dependencies, downloading the weights, and running everything locally. If you need current model or dependency details, look up the latest docs online first.