
Real app · the upload screen · the whole input
05aFrames in, your scene out
It reads the frames, not your keywords.
This screen is the whole input: one file. No prompt box, no keyword field. The model reads the actual frames, the cat's paw, the glass on the edge, and writes about that.
