My first attempt at creating a LoRA.
I pretty much played a youtube guide on training LoRAs at 2x speed, read some metadata on LoRAs from other creators, used some manually cropped screenshots from the anime, and managed to make this without having any idea what I was really doing.
Could be better, could be worse, but I think it's decent.
Generated a few images at 0.8 - 0.85 but haven't tested it much so I don't know how flexible it is.
It can do a fair approximation of several of her outfits from the anime (her yukata from the fireworks scene, her sleeveless white dress, her summer and winter school uniforms, and her casual wear) but it doesn't get any of them quite right.
Might come back to it if I ever learn what I'm actually doing, so if anyone has constructive criticism I'm all ears.
Description
Expanded the dataset and changed the network dim/alpha... still don't really know what dim/alpha specifically do, but this version seems to generate cleaner/more consistent faces. Filesize is larger though. Still trying to figure this stuff out.
I noticed while watching the anime and curating the dataset that she sometimes has sidelocks and sometimes she doesn't... you should be able to toggle them with the "sidelocks" tag in either the prompt or the negative.
Should work fine at 1.0 strength
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.