Another attempt at a "style"-type LoRA. Tested at weight 0.8. The trigger is by nefitail for V3 and nefitail for V2 & V1.
V3 was redone with a new dataset.
V2 is a somewhat stable version of the LoRA, should work with most ILXL checkpoints.
V1 is quite unstable at times. Seems to produce better results with certain checkpoints. I'd personally recommend MolKeun DeepBlue v2 EPS, but Nova Furry XL v13.5 can produce good results as well.
Description
Different dataset & different training parameters.