Version 2 (latest):

Now works perfectly at normal weights. Trained off of the same dataset as before.

===========================================================

Version 1:

LoRA intended the replicate the style of the mangaka Sakurano Ru.

Requires higher weights than most other LoRAs for the eye shape (2.0~2.3), but renders the face shape/shading at lower weights (1.3~1.8). Showcase shows the LoRA at various weights.

Make sure to include "signature" and "text" in the negative, it has a habit of rendering those even if unprompted.