← Back to context

Comment by torginus

7 hours ago

Interesting, I remembered that when trying out Stable Diffusion, once I ventured outside of the realm of anime waifus, the images ended up being so similar to existing sources, that image search could find the references.

Which is also kinda crazy since superficially there was very little similar between the 2 images, but I guess AI models used for image search converge on similar embedding than the ones used for AI generation.