This isn't necessarily a character LORA, but since this guy draws his own characters so much it can generate Cecilia, Kagari and Haiselita. It'll probably try to generate one of them anyways, there's not a lot of diversity in the dataset.
Using light particles, rimlight and maybe window shadow in the prompt can help better recreate some effects of the original images.
Used AdEMAMix as an optimizer.
More dataset rebalancing.
From my own testing this seems like it frequently comes out looking noticeably sharper and more detailed.
Go ahead and upload yours!
Your query returned no results – please try removing some filters or trying a different term.