Updated: Jul 29, 2026
style
Anima Version! I've always thought the Illustrious version was a little rough, due to a pretty limited data set I compiled while really drunk.
The Anima version to me is much better behaved and I've enjoyed seeing what it spits out.
Even then it's a little strong sometimes so I've been running it at 0.8 weight.
This style looks a bit similar to some others I've posted, but that's because it's a weighted mix of DarkfireStyle19v2, DarkfireStyle21, DarkfireStyle26, and DarkfireStyle31. I got tired of loading all those LoRA at once and balancing quality where they'd butt heads.
As seen in the Showcase images, I built in skintone separation again. I try to do that whenever I have the data for it (or I'm creating my own training images).
No tag gives fair skin, then 'tanned,' 'dark skin,' and 'very dark skin' for different shades of brown.
Additionally it's built so a girl's partner is light-skinned my default and tagging in a male will generally not have the may 'steal' the skintone from the female.
Note! I always recommend using hires fix for any LoRA I post to refine the images you like. Costs a little bit more buzz, but it substantially improves the result.