Suggested tags for consistent results:
colette loire, orange hair, medium hair, bangs, orange eyes, medium breasts, brooch, collared shirt, white shirt, long sleeves, black capelet, gold trim, double-breasted, gold buttons, pleated skirt, black skirt, zettai ryouiki, black thighhighs, loafers
Recommended settings:
DPM++ 2M
Automatic works fine (Karras tested)
Sampling Steps: 25-30
CFG Scale: 5-6
Model Weight: 0.8-1 (Tested with weight 0.8)
Clip Skip: 2
Use ADetailer to avoid blurry eyes.
NSFW generates rather well with this one.
For NSFW generations, remove the clothing tags as needed.
This LoRA works well with just about any Checkpoint that I've tried it with.
I've tested with cartoon-style through realism, and haven't found anything too egregious.
Small details change the further you get from an anime style.
I have noticed that outdated Checkpoints tend to not work very well.
The trim on the capelet can be inconsistent.
I would love to see what you create with this LoRA. If you have any feedback, I'd be be interested in hearing it.
Description
Rather versatile given the small dataset size.



