ecoprohomes/https-github.com-SkyworkAI-SkyReels-V2 ? reverse-engineered prompt

Build me a Python tool that can run SkyReels V2 locally so I can generate videos from a text prompt, or turn one image into a video. It should support the long video diffusion forcing mode too, with an easy way to pick a model checkpoint, set the output size, number of frames, and whether I want synchronous or asynchronous generation.

I’d like one simple command for each main use case, plus clear setup instructions for installing dependencies and downloading the model weights. If possible, include support for the video captioning and prompt enhancement pieces that come with the repo, so prompts can be improved before generation. Make the outputs save as normal video files in a folder I choose.

Please keep it practical and easy to run on one GPU, but also make it work for multi GPU if that is available. Look up the current model and inference docs online if you need to, and wire it up so I can try the main demo flows without having to figure out the internals.