Don't you think one of the images could be perfectly or perfectly enough encoded in that 1.9GB though? A funny example is Malevich's Red Square. Highly compress

Don't you think one of the images could be perfectly or perfectly enough encoded in that 1.9GB though? A funny example is Malevich's Red Square. Highly compressible! [0] Line drawings also can often be compressed to a polynomial. > My understanding is that any image generated by Stable Diffusion has been influenced by every single parameter of the model - so literally EVERY image in the training data has an impact on the final image. That's pretty interesting. Need to dig into the math more (lazy applications dev). [0]: https://en.wikipedia.org/wiki/Red_Square_(painting) https://en.wikipedia.org/wiki/Red_Square_(painting)