Instead of pixels, brainimg stores an image's meaning — a caption, 128² depth/edge/seg maps, and a seed — in a few KB, then > regenerates a faithful image with diffusion + ControlNets. Lossy-by-design research toy, not a JPEG replacement.
lasarux/brainimg: Instead of pixels, brainimg stores an image's meaning — a caption, 128² depth/edge/seg maps, and a seed — in a few KB, then > regenerates a faithful image with diffusion + ControlNets. Lossy-by-design research toy, not a JPEG replacement.
