UPDATE: Illustrious version now up. The dataset is the same and I have done the image training better when it came to tagging. It works great, but it might have issues with her not being able to look at the viewer when putting the "looking at viewer" prompt. So lower the weight down by .1 if you're having trouble with the that issueKim from the Illumination movie Migration. All the training was done on some 4k screencaps and 1080p screencaps from the movie. Most of the training on her was close-up shots since there weren't a lot of scenes of her, but there are some where it shows her full body, so there shouldn't be much issue here.Positive Prompts: KimMG, furry, feral, furry female, avian, duck, brown eyes, black pupils, yellow beak, blue body, white body, feathers, feathered tail, webbed feet, feathered armsAnthro version should work too as usual. However if the backgrounds that come out look weird then try lowering the weight to .9 or .8. Most of the image trained has her in a blurred background within the same area in the movie which I couldn't avoid, so see if that helps if you run into that unique issue.

Files with SHA256: 6704ee4daf85debd5dffc71c41337d4ec5d96ca0e8a14ae369fbe54d808f70be