This is B3th, my second fictional character made up from a mixture of real and AI images. Not the same technique as I used with Gw3n, and the facial consistency isn't as strong with this model.

SDXL: Looking at the training data I used for the first version, I saw what was intending from this lora and how it failed to fully realize it. The SDXL looks more like how I intended the model to look, and the training data no longer contained as many watermarks.