A while ago I discovered that a model of Stable Diffusion 2.1 that I use (Imagine v3 by vyro.ai) would consistently produce a somewhat terrifying creature when I gave it the prompt mesker (see my post A New Digital Cryptid). As I noted in that post, one digital cryptid, Loab, was found using negative prompts. That got me thinking: What does Stable Diffusion think is the opposite of the mesker? So I used mesker as a negative prompt and this is what I got:
I don't see anything consistent across those images. Clearly Stable Diffusion has a good idea what a mesker is, but doesn't have a clear idea of what a mesker isn't.
These illustrations were drawn using Stable Diffusion 2.1.
