>>105588087
I've heard that the "detail" one is actually a 2:1 merge between a model trained on 1024x and a model trained on 512x, is it true?