Heith Velvet
from
Is It Wrong to Try to Pick Up Girls in a Dungeon?
(Dungeon ni Deai wo Motomeru no wa Machigatteiru Darou ka ダンジョンに出会いを求めるのは間違っているだろうか)
This model was trained on anime screencaps from season 5 (up to Episode 11) only.
Note: This is a B-LoRA. Use the style LoRA to adjust the intensity of the anime style.Prompts:-
Trigger Words:
heith_olis, 1girl, pink hair, twintails, red eyes
Outfit:
white headwear, white dress, puffy sleeves, white gloves Optional: black thighhighs, white footwearOptional:
guro, burnt clothes pink eyes, wide-eyed, constricted pupils staff, holding staffNegative Prompts (optional):
blurry background
Weights:
Character LoRA: 1
Style LoRA: 0 for a neutral style
>=~8 for a more accurate anime style and outfit
The anime style works best with: (just in my experience)
Illustrious-XL v0.1 & NoobAI XL variants
(Feel free to experiment with the weights.)Eyes Inpainting: (for improving the quality of anime eyes, intended for use with the style LoRA)
Use ADetailer with one of these models:
> https://civitai.com/models/330727/full-eyes-detection-adetailer
> https://civitai.com/models/207406/adetailer-face-finder-furry-anime-realistic-ddetailer (this one can be buggy at times)
Width/Height: 1024 Padding Pixels: 0-64 Sampler: Euler a Steps: ~30 CFG: ~5
Denoising Strength: 0.3-0.5 (I usually do a X/Y/Z plot with a value of 0.3-0.5(+0.1) and then choose the best one.)
Positive Prompt: add both LoRA and red eyes, close-up (, optional looking tags such as looking at viewer), etc.
(^ These are the settings that work for me, not necessarily a recommendation.)Afterword:
I decided to lower the training resolution from 1024, as most anime, if not all, are produced at a lower resolution anyway. I believe this change has improved the model's quality.
Hopefully some of you will like this model.

