PKU-YuanGroup/Helios ? reverse-engineered prompt

Build me a simple app for Helios that lets me generate long videos from a text prompt, and also from an image if I upload one.

I want a clean web demo where I can type a prompt, choose a generation mode, start inference, and watch the video appear when it is done. If the project already supports different model sizes or inference options, please expose the useful ones in the UI in a friendly way, but keep it easy for a normal person to use.

Please also make sure the command line scripts still work for running inference and training, and add clear setup steps so I can install everything and get it running without guessing. If anything depends on the latest docs for video generation or GPU setup, look those up online if needed.

Keep the result practical and polished, with sensible defaults and a nice example prompt so I can try it right away.