I’ve been deprecating my old textual inversion models in favor of the (usually) superior new LoRA version. But in this instance, I feel this LoRA kinda sucks. Sure, you can eke out some interesting visuals if you frob enough digital knobs, but the quality and versatility is just not there.
Most of the shots of him are close up portrait poses, and very little variety. I chalk this up to the selection of training images (which are from the same imagery as the TI version) and my own inexperience tuning LoRAs in general.
So I leave this here as something someone might want to play with, but I will be revisiting it in the future, hoping to do an improved version of this gifted writer, weaver of dreams, visionary. Plus, actor.