It kinda sucks a bit, will retrain with more data and better captions at a lower learning rate soon, but if you want to play around a bit. This version was trained on SDXL 1.0 so should be compatible with most base models
Prompt "george costanza", can also do shirtless