Trained on every Spirit Eye Celine image I could find. You will absolutely want to put holding and holding_sword in the negative prompt, since she is holding a sword in nearly every training image. Other tips:
You can get the arm wraps by adding bandaged_arm
If your model doesn't like to render her hair pin, you should be able to get it via hair_ornament. Most models seem to pick it up, but a few have lost it.
Training images vary between silver/gray/blue hair based on lighting. You may have to add hair color to the prompt depending on what look you want.
Training images vary in eye color. I think the pink-ish color that the AI learned looks pretty good, but if you want red eyes or the more complex red-to-aqua gradient seen in some art, you'll have to prompt for that too.
Spirit Eye Celine is very asymmetric: she has a hair pin on one side, one arm is bandaged/wrapped the other has a gauntlet, etc. In my experience the AI tends to not like that and may try to add a hair pin to each side, wraps to both arms, etc. I did the best I could while training, but to get her exact look you'll probably have to do some inpainting.
The Lora looks good in a variety of models. I've tried to provide a sampling in the examples.
Description
This version uses spirit_eye_celine as the trigger, which uses a lot of tokens. If there's interest I will retrain and change the trigger plus try a few tweaks to the training tags.