V2.1: Same as V2, but trained at 768. Seems to have slightly better ear placement, but V2 is a bit more flexible with poses and full body shots.

V2: Better quality, more granular trigger word options to select ???????. Trained with base SD1.5 at 512, should work with other models, but seems to do best with realistic ones.

V1: Trained on Hassenblend using 18 images. Should work with other models.