Best girl Hokuro-Chan from Tawawa on Monday
Took a while to train this one before it actually dawned on me to train her with her monochrome images and not just the coloured in ones.
The moles may be an issue but that's just a general AI art issue, I left moles in the showcase image prompts so you can get an idea of the issues with just trying to generate them naturally. If you want the moles to be exact then you might have an easier time inpainting them. If you see images of her you'll know where they go but basically
One below her right lip
One below her left collar bone
One on the inner side of her right breast
One on the bottom of her left breast
There's also one on her left butt, but, mind the pun, there's only one image of that and I don't think autotagger detected it in the dataset so you might need to add that yourself as it might has just been destroyed as random pixels
She's got three hairstyles and outfits were added just because they exist in the dataset
Depending on which hairstyle you are choosing to generate you may need to negative prompt against the others
(TAGS)
CHARACTER - H0kur0Chn, brown eyes, brown hair, swept bangs
CHARACTER/OPTIONAL - glasses, over-rim eyewear
CHARACTER/OPTIONAL - mole under mouth, mole on neck
OPTIONAL (Default appearance) - long hair, braid, hair over shoulder
OPTIONAL (Hair down) - long hair, official alternate hairstyle
OPTIONAL (After the failed exams where she cuts her hair) - short hair
RECOMMENDED BUST SIZE - large breasts
OPTIONAL - mole on breast
OUTFIT - black serafuku, school uniform, white sailor collar, red neckerchief, black skirt, pleated skirt
OUTFIT - blue sweater, turtleneck sweater, ribbed sweater, pencil skirt
OUTFIT - straw hat, sun hat, dress
OUTFIT - black bikini, halterneck, o-ring top, o-ring bikini, side-tie bikini bottom
Trained on 37 images from the web comic
There's a lot of tags so it may add as a wall of text if you have civitai helper on your stable diffusion, sorry for this
Go ahead and upload yours!
Your query returned no results – please try removing some filters or trying a different term.