>>102590742

>>102603263
>in the preview
that's the mismatch
I'm talking doing a full denoise in just 5 steps
also, I'm using CFGStar (not CFGZeroStar) which makes a diff
(I use FreSca as well but that is a more subtle effect)

>good way to upscale
0.25 or 0.3 with euler_a at 10 steps has been good enough for me, any specific reason to go higher denoise?

>Captcha: TAMG0
the machine wants me to eat eggs lmao