JavisVerse/JavisDiT — reverse-engineered prompt

Build me a simple local app for JavisDiT++ so I can type a text prompt and generate a short video with matching sound.

I want it to handle the setup as much as possible, create the Python environment, install the needed GPU dependencies, install ffmpeg, download the JavisDiT++ weights and the required VAE models from Hugging Face, then run inference without me having to remember long terminal commands. Please add a clean Gradio style interface with a prompt box, resolution choice for 240p or 480p, aspect ratio, duration around 2 to 5 seconds, progress updates, and a place to preview and download the finished video.

Also add a batch mode where I can upload a text or csv list of prompts and get multiple generated videos saved into an outputs folder. If there are missing models or CUDA issues, show a clear friendly error with the exact fix. Look up the current project docs online if needed.