Style LoRA of Amano Kozue (天野こずえ) for SDXL.

Majority of training data is come from illustrations of Aria. Thus, this tends to generate images like Aria. If you want less Aria-ish image, add "aria_(manga)" in negative prompt.