This model is really making me question how I compile datasets. I used 4 images to make this; using my usual settings for making LoRA. Really I only used 2, but I horizontal flipped them so 2 + 2 is 4. Weird how math works.

Not at all the results I was expecting, but better than I imagined.