SkyworkAI/SkyReels-V3 ? reverse-engineered prompt

Build me a simple video generation app around this SkyReels V3 model so I can make videos from a few reference images, extend an existing video, or create a talking avatar from audio.

I want one easy command or small interface where I can pick the mode, add the image or video inputs, paste a prompt, set the length, and generate an MP4 file. It should automatically download the model the first time, but also let me point to a local model folder if I already have it. Please keep the defaults sensible for a normal person, like a short 720p result, and make it clear what each input is for.

If anything in the latest docs is unclear, look up the current usage online and match the repo’s intended workflow. Add a couple of examples so I can try it right away.