>>106362977
>>* 0.15
>why tho
just trying to stay under 4mb
>do 8x on first pass and something softer on second
I can try that. was avoiding have another model loaded but it might not be a big deal
>should have a second ksampler between
upscaling sampler is super slow. I'm already ~3min per gen, but I'll experiment a bit after it cools down later