[Asmodean]
Member
The dithering is not placebo, but as with all dithering it only works when you dither from a higher bit depth to a lower.
Thus it's currently only useful to prevent the color grading shaders and the vignette shader from creating additional banding - this however it's very effective for.
My new filmgrain shader can also be used to dither, and if you use strong enough settings it can also help hide the banding in the original game image but it cannot remove it.
To do that you would have to either smoothen the image with maybe a good bilateral filter or an anisotropic smoothing filter, and only smoothen the weak details so you don't touch the stronger details because those aren't banding.
Then you can dither the result to do debanding.
For reshade we are also hoping that it will be possible to force the games to render in R10G10B10A2 or R11G11B10, from which we can then dither to 8 , 7 or 6 bit depending on what kind of monitor the user has.
Technically we can still with the current shader dither from the games 8 bit to 7 or 6 bit, because TN monitors display in 6bit, but those monitors also have their own dithering so dithering to 6bit is usually only best for crappy 6bit monitors .. more decent TN panel monitors will look best with a dither for 7 or 8 bit.
Also 1 additional bit in my experience is not enough to make banding go away, it only lessens it - you need a least 2 bit extra to get a nice dither.
Dithering in SweetFX is now also done in the Ascii shader where it helps a lot , though I've found that I need to retune it for the ascii effect. The current dithering I'm using there works fairly well but I think I can do better. Right now though I'm busy with other stuff though.
Hey, c'mon now. I really need to be more careful with the whole perceiving text stuff as literal. What I meant was - adding conventional, external PP dithering isn't going to magically resolve banding of the game itself (prime example is skybox compression).
It's only going to help with banding introduced with other post effects preceding the dither. As I also mentioned.
What you're describing above sounds interesting, though. I can see it working. Saying that - there's a difference between "resolving" the various banding types, and simply masking them, with the effect.
I like the idea of shifting RT bits from alpha to rgb. I've had some stuff along the same lines as this, that I've been experimenting with previously, to overcome/work around other 'issues'. I get bored quickly though, with little motivation working alone on stuff. When it takes too long with little results. I typically move on. (until I get the itch to check it out again) =)