SD-latent-mirroring
SD-latent-mirroring copied to clipboard
dfaker
Doesn't seem to have an effect on the 768 model
Hi there, Been trying to use this with the sd 2.0 model and am not getting much if any effect. Am I doing something wrong?
Oh does it only work with some samplers? I changed samplers and got a couple to work but only a couple.
That makes sense, I'll see what I can find in terms of determining if it's a second his-res pass.
Not guaranteed to work equally well with all samplers makes sense for DPM++ as it's a higher order sampler and is using prediction which you're frustrating by messing with the latents.
This makes a lot of sense because I figured out how to make some of them sort of work but it required HUGE numbers of steps.
This is a weird thing you helped make, thanks!
Which samplers would you suggest as being the best while you work on this?
I've never attempted to rate them on how well they deal with the mirroring, Euler Ancestral has always performed well for testing.
Doesn't seem to do anything. Would be good to provide some tutorial in the description because I used the same values, but nothing happens.
