(v2.0)
Updated LoRA with a bigger dataset and trained on NAI. Should be more consistent generating images of Noihara Himari, but it can also generate other characters from the show very closely.
The trigger word for this version is promotionalart but it can work without it.
(v1.0)
I had first used SD 1.5 Pruned for training but It didn't replicate the art style that I wanted, then I decided to go with Lykon's AnyLoRA - Checkpoint (Pruned) for training which gave me just what I was looking for.
I tested mostly on AnythingV5 (works with other Anime style models) with 0.55 - 0.65 weight which gives me a very decent depiction of the character along with the art style, while also allowing some flexibility to change clothes poses and backgrounds, however since the training was done with only one image, the AI will attempt to generate something similar to the base image used for training, at anything above 0.8 weight, the AI will generate a similar pose to the base image more often while still allowing different clothing. No trigger word needed, but adding OmaHimaStyle seems to add a bit of extra detail (maybe is just extra noise, idk).
This was mainly made to generate Noihara Himari images but surprisingly It can also be used as style LoRA when combined with other models, however this LoRA sems to start breaking with school uniforms when above 0.8 weight, likely because of the base image having a school uniform idk, I just wanted something quick to try out, I'm quite impatient. Enjoy (I guess?) but it can be fixed by adding more weight to specific tags.
Basic prompt: Solo, 1girl, portrait, Black hair, purple eyes, very long hair, high ponytail, long ponytail, hair bow, pink bow, big hair, red shirt, sailor shirt, bangs, sidelocks, <lora:OmaHimaStyle:1>,