General:
Someone reminded me of this old rpg maker game, it was a nsfw League of Legends parody. I tried making a test character lora with it to see if it was possible. With some careful upscaling and fixing ugly transparency artifacts it worked out quite well (despite limited dataset).
Tags and information
Use tags for base:
nn_riv3n, 1girl, white hair, red eyes, facial mark, lips, folded ponytail
+
Outfit 1:
chest sarashi, bandeau, white skirt, belt
Outfit 2:
rabbit ears, playboy bunny, purple leotard, detached collar, necktie, wrist cuffs, thighhighs
Outfit 3:
maid, maid headdress, detached collar, detached sleeves, wrist cuffs, dress, waist apron
As the loras often are, it's not perfect. It has the style in it partially, maybe I'll try doing the style separately too later. As for the other characters, it won't be a priority right now, but if there's significant interest in this one, then I'll probably do it.
Support:
If you like what you find here, please consider making a donation https://ko-fi.com/hat0ri, thanks!
Description
Small thing to note. I trained two versions basically - first one only had one outfit, for the second one I added two more, but the tagger I used didn't put the "holding" tag in captions for two images where she was holding a sword (and I forgot to do it manually), so if it affects generation, put this into negative prompt:
holding, broken sword