This one came out a little weird, but compared to what I was getting by hand, I think this is much closer. For this one I used a fancy program to generate text from the original image, then I used that prompt to make variations and ran that back through stable diffusion. It isn't really that close, but this was a particularly hard character with lots of details.
