trained on the same images as my last 1.5 version, just for SDXL. I haven't tested it so I don't even know if it works well.
Liminal.safetensors