FlashML-org/FreeVideo ? reverse-engineered prompt

Build me a desktop app and local workflow for generating videos on my own computer with FreeVideo, so I can type a prompt, optionally add a first frame, last frame, image, video, or audio reference, and get a video back without needing cloud processing.

It should work on Windows, macOS, and Linux, open into a simple browser based workspace, and feel easy for someone who is not technical. I want an install flow that can reuse an existing ComfyUI setup, download missing models automatically, and also support offline installation when needed. Make it smart about hardware, so it can run on weaker GPUs and use whatever acceleration path makes sense for the machine, with lower memory use and sensible quality options like Light, Medium, High, and Max.

Also include a clean way to preview results first, keep a history of past generations, and let me export the finished video along with the prompt and settings. If needed, look up current docs online and make the setup as smooth as possible.