>>75259625feels like it's really effectively a cheap version of doing the AutismMix(base)->Animagine(refiner) but I'll have to give it a test side-by-side to see how well it actually works. Prompt editing is perfectly capable of doing cool things with the refiner methodology, but the differences in tagging between the two models is enough to make me believe the refiner method should be better, even if it takes longer and requires more handholding.
