Most of the images from the training set were generated using pony, incase, all Disney princesses, and expressiveh. There are few other images.
Trying_again.safetensors