It took me a while to work on the next version, but here we finally are! After beefing up the dataset with a lot more generations, adding more checkpoints to the equation and going through multiple failed attempts I managed to get a slightly better result. It's still far from ideal, but at least it gives somewhat decent results now. Or at least it did for me, I can't wait to hear what you all think of it!

Model was trained on images generated by me using Nova AnimeXL v18.0, Nova Furry XL v17.0, PlantMilk model Walnut and WAI-illustrious-SDXL v. 17.0.

Trigger words: laur-wolf, heterochromia,

Recommended weight 0.8-1

I've made example generations using 30 steps, but experience may vary for each person so this one is just an advice and not recommendation.

Thank you in advance for checking it out!