Originally this was going to be two different LoRAs, but I wanted to see if training both datasets at once with tags for both similarities and differences would help get the concepts/styles to work better. I think maybe it did.