I'm sure we'll have a wonderful time, snookums!
I figured I might as well continue the Monster Musume loras for a while longer, and I figured that Tionishia would be the most suitable character to do next. Mostly because we all need more onis in our lives.
Main tags: mmTionishia, red eyes, platinum hair, long hair, parted bangs tan skin, monster girl, oni, giantess, single horn, huge breasts
MON uniform: sleeveless long coat, yellow coat, black sleeves, pauldrons, black pants, silver elbow pads, silver knee pads, utility belt, black combat boots, silver gauntlets, long-sleeved black undershirt, yellow shoulder gloves, yellow thighhighs
Casual outfit: tan skirt, white shirt, frilled skirt, long skirt, yellow jacket, frills, black footwear, cleavage, beret
Underwear: underwear, pink bra, pink panties, frilled bra, frilled panties
Swimsuit: swimsuit, bikini, micro bikini, pink bikini, midriff, navel, hair flower, scrunchie, skirt
Pajamas: sweater, pink sweater, striped sweater, sweater dress, off shoulders, bare shoulders, panties, white panties
Workout: track jacket, spats, white skirt, pleated skirt
Suit: suit, black suit, black jacket, black skirt, white shirt, black tie, pantyhose, black shoes
Artstyle tags (Anima only): @okayado, anime screenshot, game cg
Strength-wise, this lora should work best at 0.8, but it certainly shouldn't have any problems at 1. Same goes for Anima, I suppose. I also strongly recommend that you put the tag tan lines in negative prompt.
Description
Tionisha for Anima. Diversified the dataset a little to include Okayado's official art into the mix, and as such the dataset is trained with three distinct styles, anime style, manga style and game style, reminiscent of the old monster musume gacha. From my test generations, everything seems to work just as fine as the initial Illustrious release with all outfits included, save for her uniform, which ended up looking completely wrong with the original tags. That is something I'll specifically have to rectify for future releases, so for now I'm just gonna provide the tags that did work to replicate it. Beyond that, it learned the actual character just fine, and you probably will get some good results with the included artstyle tags as well. 0.7 should be the recommended weight, but 1 can work if you're not gonna use artstyles not included in the training.