The is LoRA using LoCon as convolution network, which can be loaded by using LoCon extension:GitHub - KohakuBlueleaf/a1111-sd-webui-locon
Install it through extension “Install from URL”, and you can use locon model as LoRA. (Put safetensor in LoRA folder)
You can learn more about LoCon from here.
Please use this with respect for the original content, DO NOT publish any NSFW content.
For more accurate result, please use the following prompt for each outfit
Racing outfit
fine motion \(umamusume\), shoubufuku, white jacket, black bow, button, open clothes, green shirt, long sleeves, white gloves, black skirt, white socks, white footwear
Wedding dress
fine motion \(umamusume\), tiara, bridal veil, yellow hair ribbon, lace sleeves, long wedding dress, lace dress, high heels
Casual Wear
fine motion \(umamusume\), fudangi, white shirt, long sleeves, belt, long skirt, green skirt, sandals
School uniform
fine motion \(umamusume\), natsufuku, sailor collar, puffy sleeves, short sleeves, purple bow, tracen school uniform, serafuku, white skirt, pleated skirt, white thighhighs, brown footwear
Prompt is broken down to more complicated one because condensed version was unable to reproduce certain clothing details properly.
*Use " socks ", "sandals", "thighhighs", "footwear" only while drawing lower body or full body, or it will force AI to draw in full body scale.
**Use "from behind", "ponytail" in negative prompt may help
For unknown reason AI recognize hair bun as ponytail (mixed with horse tail?), add “short hair” in positive prompt while “ponytail” in negative prompt will help.
Racing outfit may generate white skirt (mixed with jacket) sometimes, add "white skirt" in negative prompt will help.
School uniform may mix with racing outfit sometimes, add "shobufuku" in negative prompt if necessary.
Adding "horse tail" will generate tail but dont expect on right place.
Again, Please use this with respect for the original content, DO NOT publish any NSFW content.
(If you want her to eat ramen, try to use inpaint separately is recommended)