@anime_outdoor
Literally an inevitable result of the diffusion model.
A difusion generators work by adding noise, and then trying to repair it by using prompts as an instruction. So, even if you tell it to literally return the same picture, it will take it, add noise, and then tries to correct it to "look like original" without knowing what the original actually looked like.
@anime_outdoor
Hair over the ear is not more suggestive.
2 necklaces are not more suggestive than one
I know what you think you see, but can you please re-create the experiment with a fully civilized image, just to see If my understanding of AI dumping things onto a diffusion model is correct?