My first attempt at making a textual embedding model for Stable Diffusion.

I found that it works best with an emphasis between .6 and .8.

I generated 9 images for 3 prompts and chose the best 3 of each of them for his comparison. They certainly can be better quality, but I wanted to focus on relatively easy to obtain, and relatively common results

This model works pretty well with ProtoGen models, but will probably work with Anything and SD 1.5 based models as well.

The negative prompt for all of these images is my default negative prompt for portraits