This is a LORA trained on the style of Kan Liu using animefull-latest. It excels at water effects at even low strengths, and can be used to good effect all the way up to roughly 0.9.

This should work decently well on most anime models and was trained on CLIP skip 2.

I'm taking a break from this for a little bit, but I plan to take another crack at things like limited clothing and hair soonish. I have spotted some unprompted fire issues in v2. All images in the dataset that featured fire were tagged accordingly, and I've found some success with the negative prompt "fire, embers" to keep things cool.