The object of the exercise here was to see if I could make a competent lora with a synthetic dataset derived from a single image. The answer is, "yeah, kinda." Style and certain features are baked in pretty hard, so if you want her to have anything other than blue bits, good luck, and my apologies in advance. Also struggles really, really hard with doing an image from behind. Not really sure what's up with that.

Tag: F3mw0lf

Really that's the only one you need, but certain reinforcement tags can help as well: black fur, solid eyes, glowing eyes, blue eyes, long tail