Same as the other one but trained with Ip Noise Gamma supposedly it would require less steps and produce better images. It looks exactly the same to me but get both of them and compare.










Warning: Loli only loving deviants beware, this is liable to create big ????? due to using an older Reimu in the dataset.
Bullet hell? What is that? Is it edible? Truth is I mostly know Touhou through the eraTouhou ???? text adventure games and through doujins. Never played the original games nor do i plan to.
Regardless! Here we have our favorite miser miko of the barrier. I did my best to go for a more mature Reimu, there's enough canon or loli versions around, hell even OG NAI can more or less cope.
works fine with: <lora:ReimuV1:0.7> hakureireimu,
I added an alternate version trained with ip noise gamma at .1. I can confirm it doesn't train "faster" otherwise it would have burnt and i would need to publish an earlier epoch. As for looking better... I dunno. Check them and compare I guess.
The ip noise gamma one seems mostly the same try it with: <lora:ReimuV3:0.7> hakureireimu,
Well she is Reimu so I mostly found images of her miko variant outfit, I did manage to get a lot of skirt differences so I split into skirt and miniskirt versions:
nontraditional_miko_Yellow_ascot_Red_crop_top_Collared_shirt_Detached_sleeves_Ribbon_trim_Red_miniskirt_white_thighhighs: Miniskirt version of her outfit.
nontraditional_miko_Yellow_ascot_Red_crop_top_Collared_shirt_Detached_sleeves_Ribbon_trim_Red_skirt: medium to long skirt version of her outfit.
I'll add the full dataset together with the tag summary so people can experiment.
Same as the other one but trained with Ip Noise Gamma supposedly it would require less steps and produce better images. It looks exactly the same to me but get both of them and compare.
1. Os direitos dos modelos repostados pertencem aos criadores originais.
2. Criadores originais que desejam reivindicar seus modelos devem contatar a equipe do SeaArt AI pelos canais oficiais. Clique para reivindicar
