Goal:
This stable diffusion prompt gave me a lot of insight into what Stable Diffusion will do with a prompt that involves "a child". I wasn't sure how to describe the Scavo Technomancer other than as a "small wolf child", which kind of worked, but I was getting a lot of really creepy, super realistic children photos in like... Halloween costumes of wolves. Not exactly what I wanted, but every once in a while I'd get a wolf/human hybrid that almost looked good enough. It seems that the internet dataset the algorithm is trained on is full of children in costumes, and all photorealistic. It really seems to bias any prompt photo with "child" towards a photorealistic version of one, even when the rest of the prompt is full of fantasy characters or descriptions. I think there were a few that all looked like "the best". Getting any kind of gadget to be in its hand was almost non-existent even after 20k+ generations. And goggles were kind of hard to get in conjunction with the correct facial features.
And of course there were more failures that I could possibly count. These were the more interesting of the failures.
Feel free to suggest a character for my next one!