Sign In

Satsuki Kitaoji - Ichigo 100%

23
79
0
0
Verified:
SafeTensor
Type
LoRA
Stats
79
0
Reviews
Published
Oct 21, 2024
Base Model
SD 1.5
Training
Epochs: 8
Usage Tips
Clip Skip: 2
Strength: 0.7
Trigger Words
kitaooji_satsuki_high_ponytail
Training Images
Download
Hash
AutoV2
98C662AC06
default creator card background decoration
knxo's Avatar
knxo

Mizuki Kawashita is basically the "We have Katsura at home". I first read ichigo 100% eons ago a bit after reading I''s. She more or less follows Katsura's style but the story is crappier, the characters lamer and now collecting the dataset from the manga the boobsize and other characteristics are as consistent as poorly set jello.

In my mind when i started, i was thinking "I remember Satsuki being the best of the bunch" after giving the manga a spin to crop and clean the images it turned into "Holy Ahura Mazda on stick, all of them are terrible!" So it is less that she is the best of the bunch and much more that she is the least worse of the bunch.

It might be the trying to apply some sort of real world logic to a harem manga but in retrospect now that i am a cynical old man instead of a teenager, everything is extremely cringe. In fact checking it again after so many years it gave me Schooldays vibes by which I mean everyone needs a bitch slap.

Let's not talk about the anime, the animation quality is about 3 notches below poor, needless to say only a couple anime images made it into the dataset, less than 5 possibly only 1.

works ok with: <lora:SatsukiV1:0.7> kitaooji_satsuki_high_ponytail,

I excluded from the dataset her lose hair version and twin tails one because there's just a handful images in the manga and her iconic look is with her ponytail.

Outfitwise, she only uses typical school uniforms white and blue with red ribbon and navy blue with a red ribbon. so I didn't bother as all models are able to spit them out. Other outfits only appear a handful of time so not really worth i trying to train them

As this is made with only the manga and some fanart, i won't be adding the dataset instead i will add the tag summary for people to experiment.