I wonder if it's possible to assign a unique id to each image with a fingerprint, then feed it to an AI model when training. Then, on inference, be able to estimate the weights of each reference image used to produce an output. I bet they'd be a lot but still useful output.

The open web is dead, long live the open web. ⌐◨-◨

Building /warpcast and /farcaster

— gabrielayuso.com

Have you been following Holly Herndon and May Dryhurst’s spawn/holly+ project? They’re doing something like this for audio inputs to AI