
This LORA was trained primarily using Koikatsu screenshots and some AI generated images. I've attempted to curtail style burn. But I think as a result certain character features don't work consistently on base Illustrious and may be more reliable on merges. I don't know why. Outfit accuracy is also hit or miss.
It uses the following tokens for the character and various outfits:
-
kanan zebel, medium hair, demon horns/two side up/twintails, multicolor eyes, tsurime, pointy ears
-
officer, peaked cap, black dress, mauve pantyhose, belt buckle, half gloves, high heels
-
casual, sleeveless sweater, layered suspender skirt, high heel boots, black pantyhose, heart necklace
-
dress shirt uniform, blue pleated skirt, short/long sleeves, knee socks/brown pantyhose, grey footwear, striped bow
-
vest uniform, grey vest, blue pleated skirt, short/long sleeves, knee socks/brown pantyhose, grey footwear, striped bow
-
jacket uniform, grey pleated skirt, navy jacket, dress shirt/grey vest, knee socks/brown pantyhose, grey footwear, striped bow
-
waitress, blue dress, headdress, striped bow, ribbon thighhighs, waist apron, back bow, mary janes
-
open clothes waitress, headdress, striped bow, ribbon thighhighs, waist apron, back bow, unbuttoned dress, pink bra, mary janes
It uses the following style tokens:
-
koikatsu 3d
描述:
训练词语:
名称: kanan illustrious 129.safetensors
大小 (KB): 144177
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success