I decided to do a big mashup of 8 of my recent SDXL models just to see how it would come out, and I got this. I wouldn't call the style super distinctive, but it's a nice anime model that renders at high detail and can do hands and bodies without any serious flaws. It's also quite good at landscapes and architecture, and follows short prompts well. I imagine it will be a good base for further training and mixing.See <a rel="ugc" href="https://civarchive.com/articles/3148/envys-cauldron-05-my-image-generation-settings">this article</a> for details about my generation settings.

EnvyMegaMixXL01

v2 is better than v1 overall, but it took a bit of a quality hit in some ways, particularly pertaining to anatomy. This v3 release shores that up a bit while retaining improvements to details. In particular, I'm able to turn off adetailer and faces are still clean and sharp.

Files with SHA256: 6e5486579eb397499cf93d9280a46a5047c03fe00f3736e3e67fea81d653dc7c