LoRA for Fremy. Recommended weight 1
Optional: forehead scar, heterochromia, pink eye, blue eye
Also, like 95% of the dataset was the same outfit, so not sure how flexible it is with other stuff, though from my testing it's surprisingly good. Results may vary though :/
First time trying out Dadapt training method, used 75 images, 1 repeat, 250 epochs, 8 batch @ 512x512. Took the 190th epoch, turned out pretty ok, though if you reduce the weight below 1 it seems to quickly lose resemblance to the character. But it does seem to be pretty compatible with anything at 1. May try to overfit little more next time