I gave this a comparison (same prompts and seeds) to my own version of a Vanilla Cookies IL style model I trained a few days ago. This one has a much stronger effect, which is going to be better for many applications. I'll definitely be using it some. With my own version, I included artificial data in a successful (I think) attempt to capture the faces while excluding a lot of the overall style, so they're very different in purpose. I'm glad we have a more pure style option for IL now. Thanks for this!