They’re made using Stable Diffusion and the controlnet model. You use a black & white image to bias SD toward the design you want as the hidden message in the final image. If you saw the renaissance village spiral image that’s been all over the internet for a couple of weeks these are made the same way:
They’re GPU time intensive so I don’t know of any free site that’ll make them for you, most people just fire up stable diffusion locally. I’m sure someone will come up with a minified model that’s cheap enough to do that eventually but I don’t think it exists yet.
Add comment