Fuck, this doesn't look right. The first time I trained a lora, I didn't know what I was doing and used Pony as the base model and things turned out great. While the Epochs were being generated, I was getting relevant preview images to what I was training.

Now I know a bit more and chose to train with Illustrious because that's the model I like using, but the previews are so generic, these don't resemble the character I took pictures of at all and I took (and edited) really clean reference images and the auto-tagger was on point with the descriptions. Fuck, I think I botched it somehow. Either I fucked up somewhere or Illustrious-trained LORAs don't create good previews.