SDAgnis, red eyes, expressionless,, ponytail, hair ribbon,, ,
https://civitai.com/models/701022/agnis-7-outfits-the-scarlet-demonslayer
*This model is a DoRA. If it doesn't work, you need to update Stable Diffusion. Put it in your lora folder as usual.
Here's a model of the Agnis, the main character from the RPGMaker game The Scarlet Demonslayer (aka. Akagami no Kishin). Agnis is completely expressionless / emotionless in-game, but you can make her smile with this model if you really want to... I guess... 😒
Generation settings 1.0 strength for general use.
0.7-0.9 with weaker style loras.
Add source_anime in positive and 3d in negative to follow the original style more.
Base appearance Main tags: SDAgnis, red eyes, expressionless,
ponytail, hair ribbon,
OR hair down,
Outfits Leotard: black leotard, covered navel, sheer covered cleavage, sleeveless, thigh boots, highleg leotard, skin tight, lace trim,
Casual outfit: white sleeveless turtleneck, thighhighs, miniskirt, black skirt, garter straps, midriff, cleavage cutout, side slit, bare shoulders, taut shirt, high heel boots, optional: bra lines,
Bunny waitress: red top, black bunnysuit, underboob, wrist cuffs, shirt collar, necktie, pantyhose, bare shoulders, sleeveless,
Exhibitionist: transparent brown top, nipple cutout, midriff, microskirt, black skirt, garter belt, garter straps, thighhighs, panties,
Harem dancer: harem dancer outfit, mouth veil, hair veil, jewelry, see-through, underwear, covered nipples,
Underwear: underwear, lingerie, black lace bra, black panties, garter belt, thighhighs,
Armor: red armor, mask, gem, pauldrons, cape, dragon horns, fake horns, hidden eyes, white pants, armored boots, Don't prompt for "red eyes" and "hidden eyes" at the same time
Known issues Biased towards white/back/simple backgrounds unless you prompt for something else.
Slight bias towards standing with her arms at her sides (i.e. the pose she has for like half the images in-game).
The waitress, exhibitionist and dancer outfits had few training images, so they're not as consistent as I'd like.
Overall I'm surprised at how flexible the model is considering that the dataset is 100% images ripped from the game files with some backgrounds slapped on. And upscaled, of course.
If you like the model, upload some images!
