>>2842759Creator simply claimed Euler a @ 25steps is ideal with no explanation as to why. I was struggling to get good gens from pony like everyone else so I switched to it and the gens improved so I stuck with it even though I do not prefer ancestral samplers cause no converging. I prefer dpm++ 2m cause it converges around 20-30 steps with good quality. Someone did some interesting research on the samplers here:
https://stable-diffusion-art.com/samplers/ not sure how accurate it is but I learned a good bit from it.
