The model I think is fairly self explanatory.

Works in bizarre ways due to the limited dataset but I think I solved it by having a strength of .7 for the lora and a strength of umeko_vampire:.4 for the primary tag. Not entirely sure why this one got complicated as opposed to reiko as they're fairly similar in design, but I guess it's all relatively a mystery at the end of the day.

Previews made with autismmix confetti.