Character model based on 34 (pony) / 38 (noobAI) images of Indiana from gacha game Azur Lane. Namely the official splash arts, a few official artworks, some screenshots from the music video, and the small handful of fan arts.
Usage
Basic
Works well anywhere between a strength of 0.8 to 1.0.
Base trigger is indiana
Specifying dark skin shouldn't be necessary, but may help improve the result.
The face scar is tagged separately in case you don't want to include it, but can be inconsistent.
Higher strengths (0.9-1.0) will more accurately capture the face scar but may lead to the base outfits bleeding into the output.
Outfits
Outfit components are tagged separately so you can choose what to include.
Base skin:
long coat, fur coat, necktie between breasts, harness, cleavage, midriff, black chaps, high-waisted thong, black boots'High Noon Arrest' skin:
cowboy hat, blue cape, tassels, black bra, cleavage, under boob, belt, hotpants, cutoffs, chaps, crotchless legwear, fingerless gloves, brown boots
Extra
In the interest of openness and community-helping, I have uploaded the dataset along with the model, this includes: images, captions, and the training json (for kohya_ss).
Description
Increased dataset size to 34 images (+12)
Upscaled the lower resolution images
Wider dataset means overall style is a lot less affected now, and the upscaled images means that the face is reproduced a lot better. face scar consistency is a little improved, but not much.