Character consistency tests

Seems like i haven't posted on here in forever. Work exploded, and I accidently became head of ai at the studio I work at. Awesome, but also time consuming. I've developed a lot to of things I just can't show you. But I finally got a little time to myself, and wanted to try out the lora trainer on - site, so i trained a lora with an old dataset, to see how it did in SDXL vs 1.5.

It did pretty darn good, and i'm getting some great consistency out of if using Pony diffusion, which does expressions a lot better than most of the models i've used. Because of the restrictions i've set on our use of AI at work, I can't use Pony, so i haven't really played with it before now. Looking forward to making more images of this character and increasing her dataset, and retraining her again.