trained on IllustriousXL, will (probably) not work on any Pony/Animagine-based model !
Port of the PonyXL version using the same dataset. Obviously weaker at things it doesn't know (like NSFW or landscape only).
As always with styles, the result will depend heavily on weight, knowledge, and checkpoint its used with.
There is NO trigger word, it works from just being added to the prompt.
It has been tested from 0.4 strength to 1.6 strength, recommended strength is 1.0
The game's characters weren't tagged in.
dataset: 220 images from the game(99% characters splashes) cropped to different views.
Huge thanks to @PotatCat for bearing with my single braincell(which is in fact my only cell as blobs are unicellular organisms) and patiently guiding me on how to cut unneeded portions of the lora. I can't thank them enough.