V1:

curious about how good is lora at adapting blender rendered models, and so I trained a model over a 3d koikatsu artist, specifically selen and doki.

V2:

Much much better version of the style, use 3d to get the style, and much better anatomy, fingers, and feet. I still recommend using adetailer to get the eyes right, like the samples, particularly for dokibird, but for selen, its pretty decent without.