V1: Very flat colors, lineart, first attempt, used bad dataset and bad tagging, 25 epochs and 9,175 steps.

V2: Fixed dataset+tagging, cleaner results. Resulted in half original dataset size so its now 35 epochs and 3360 steps.

V3: Used 51 epochs and 9200 steps here, but is it the best result? I dont know, thats why I decided to release all three at the same time!

If you have lora requests (SD1.5, SDXL or Pony) post it in comments/discussion! (artist styles)