New training based on 200 more images than before; trained primarily on Base Anima, but it will likely work with any Blender-based model (I use "aesthetic 1.0" for my images—you can find the data below each image).