Contessa is an original character of Icymasamune, from an as-of still pretty short series. I liked her design, as complicated and thus dreadful for AI to learn as it may be.

May need a detailer but if you know how to use it you probably don't need told that. Otherwise, it's pretty friendly to inpainting. Can always just remove the red eyeliner tag.

Keep in mind that using the gold lotus accent tag when generating from behind or other positions where it won't be visible can cause some AI wonk to happen. Had to train it as its specific thing, since training it without causes it to always come out really weird. I swear just having the "gigantic breasts" tag always makes everything much more unstable on every model.

Also tried to get it to remember her outfit as one tag and it just didn't work out. This is actually like my 20th personal version in testing things, and decided this one was workable enough. Might try to update it in the future, might not.

Trained on Illustrious v2.0 with some 150 images. Tried to focus on consistent small details but you know how AI generation is. Funnily, this character is never seen completely from the front in the original work.