V2 uses a narrowed down dataset which I think made the style a bit more accurate.
V2: No activation tag necessary
Description
FAQ
Comments (4)
it's a good style but looks nothing like xDomiKami
Hmm, it was trained on about 300 of their most recent images. Perhaps it's overtrained?
EDododoJul 18, 2026
@gothwomenenjoyer i actually trained RedIlus's style which is similar to xDomiKami using 300 images too and it turned out pretty bad. I have no problem with 300 images on illustrious lol. idk whats the problem.
@EDododo Interesting. I may try to narrow down narrow it down to like 150 and see if the results are better