Here's Sekirei's best girl. If you find yourself disliking tsundere girls you must remember that everything is the fault of those whimsy male leads that don't properly discipline them! Remember to spank your tsundere! ...unless you like to be spanked yourself then I guess you don't.V3Ugh... This was old, even the retrain was old. Long before bucketing was proven. So the opriginal size dataset was gone and only the resized one remained. So I pretty much had gather it and clean it from zero. This time it is mostly trained on the manga.Works fine with: lora:TsukiumiV4:0.7 tsukiumi,I was able to isolate 2 outfits and her twin tail version. In the series there's also a blue dress but there weren't enough images to be worth it. The outfits are:twintails: for her twintails guise.blue_short_dress_maid_headdress_detached_collar_waist_apron_blue_bowtie_puffy_short_sleeves_cleavage_white_back_bow: Her blue maid outfit.black_dress_white_dress_underbust_center_opening_juliet_sleeves_overskirt_frilled_choker_shoulder_strap: Her classic outfit.thigh_strap: for her OG outfit garter strap thingies.No dataset for you this time as it is largely trained on the anime and manga. So only the tag summary.V2:Argggghhh! My old dataset was dirty and not in the good way. I mean it is understandable for my second ever LORA but fuck you past me!Works fine with: lora:TsukiumiV2:0.7 tsukiumi,I was able to isolate 2 outfits and her twin tail version. In the series there's also a blue dress but there weren't enough images to be worth it. The outfits are:twintails_Tsukiumi: for her twintails guise.Tsukiumi_Blue_Maid_Dress: Her blue maid outfit.Tsukiumi_Black_White_Dress: Her classic outfit.As Always I will add the tag summary inside the dataset so people can experiment.V1:Her hair color seems to deviate to darker tones, maybe becauser i used some manga for the training so use blonde_hair if you want to make sure.Had good results with: lora:Tsukiumi:.4, (Tsukiumi), blonde_hair, big_breastsMost loras seems to do good with .4 weight and double weight for the trigger.
