>>86065761https://files.catbox.moe/u2jcc2.pnghttps://files.catbox.moe/hfb9uj.pngOkay, final comparison for today. Honestly both heun and uni_pc were so similar to euler that I thought I had broken something, but no, it seems they are just that similar in result, cause euler_dy_negative still results in something different(surprisingly so, I think the smea/dy samplers are the ones I should test the most, cause they result in such different images.)
Dunno if I should tag the anon that wanted naked mio sleeping, cause these weren't really meant for him, but I just happened to do his request by coincidence.
G'night /vtai/
>>>86066346Yeah, I also like it. It looks blurry, but not in a bad way.
>>>86067909>so different samplers/scheduler have their own cfg that worksYeah, biggest difference is obviously between the cfg++ and base samplers, but seems like all of them have their sweetspot. Especially on Illust, which seems to be especially prone to washing out or frying at the "wrong" cfg.
>smeayeah, that and the dy samplers is next on my list. Could definitely be an installation issue, but could also just be that they are implemented weirdly, since it was just one guy's assumption of what the NAI samplers do, as far as I'm aware.
