trained on Pony Diffusion V6 XL, will (probably) not work on any Animagine-based model !
I had the images laying around so I tried making it. Works quite well, obviously weaker at things it doesn't know (like NSFW or landscape only).
All preview images were made using autismMix Confetti because it has the "weakest" base artstyle of the pony-based models, if you use any other pony-based checkpoint (like Incursio's meme diffusion or Mfcg) you'll have some influence from the models.
There is NO trigger word, it works from just being added to the prompt.
it can work on pretty much any
It has been tested from 0.4 strength to 1.6 strength, recommended strength is 1.0-1.2
The game's characters weren't tagged in.
Funnily enough it works quite poorly with the characters LoRA due to the quality of their datasets giving them a style that competes with the LoRA.
dataset: 212 images from the game(99% characters splashes) cropped to different views.
Q: Why is it not a lycoris?
A: Because I forgot to switch from lora to lyco but it's a LoRA with the same values as lyco.
Q: will you train a lyco version?
A: maybe, if I have the time.
model/lora style influence example: