Higher quality with fl2va than with ref2va, bith with refs - This can be deleted - mistake
#128
by juang3d - opened
Edit: my mistake, this can be deleted, it's true that I see higher quality with fl2va, but it's not so big like in this example.
I tested the fl2va with references and turns out it works quite well, and the quality is way bigger than the ref2va model
In the example you see the EXACT same workflow, only changing the model, the first image is the ref2va output and the second and the third are the fl2va outputs.
What's happening? why do we get such a worse quality with the ref2va model?
juang3d changed discussion title from BIG Quality difference between ref2va and f2va, both with reference latents to This can be deleted - mistake
juang3d changed discussion title from This can be deleted - mistake to Higher quality with fl2va than with ref2va, bith with refs - This can be deleted - mistake