flashml-org/freevideo ? reverse-engineered prompt

Build me a local video generation app that runs on a normal consumer GPU and still works on lower memory machines, ideally around 8 GB of VRAM and 16 GB of RAM.

I want it to let me make short videos from a text prompt, and also support using a first and last frame, plus image, video, or audio references if I have them. It should adapt itself to the hardware it finds, so it can choose the best acceleration path automatically and keep memory use low.

Please make it work nicely inside ComfyUI with a simple creative workspace, support batch generation, keep a history of past creations, and let me switch to a more advanced node view when I want to tweak things like LoRAs. It should also have a Windows launcher that can set everything up for me, reuse existing models if they are already installed, and support offline installation. On Linux, a command line way to generate a video from a prompt file would be great.