Trained on the artbook Soul Eater Soul Art by Atsushi Ohkubo.
Strength of 1 is usually fine.
Not trained on any triggers, but using Anima's own tags @ohkubo atsushi, traditional media, can help reinforce the style.
This LoRA does have a few small issues, I might end up fixing up the dataset a bit and trying again with a second version later, but it does a good enough job capturing the art style. The tag (wrist cuffs) may need to be added to your negative prompt at a high weight, that part of Maka Albarn's outfit may have bled into the LoRA a bit too much
Update: Most issues fixed in V2. wrist cuffs can still bleed through, but overall much more stable after cleaning up the dataset captioning.
Description
Cleaned up dataset captioning. Improved stability, text, lower frequency of unprompted wrist cuffs, better moon. Slightly weaker than v1, but not too noticeable






