Thanks for the in depth response! I kinda missed it and only just noticed.

I’ve tried a few refined models by now and those noticeably improve my results, but sadly I haven’t managed to get SDXL models to optimize and/or get them working yet (running it with Olive on AMD).