My first LORA submission; I was having trouble getting some reasonable generations with specific models, so I tried making one of my own.

Tag: <lora:Koyan-dama-2> (recommended weight: 0.5-0.8)

Activation: koyan-la1

Recommended tokens: black bodysuit, cleavage:1, partially unzipped, pink hair, hip vents

Training Model: AnyLora

I find this works well at medium-low step counts (25-35 on Anylora V5), and really high step counts (100~ on most models), it really depends on how temperamental the checkpoint you're using is.

Please leave feedback so any future LORAs I decide to produce can benefit from the advice.

Thanks!