IRRELEVANT
So I downloaded my own fork from scratch again and used the 1943 preset - changed to Presets-ng/Monitor-Ambilight-immersive.slangp Both FBNeo and MAME2003plus have the same result as enclosed. Means that I have to double check with Windows. I truly hope that this is not MacOS specific.
Third edit after thorough testing: It’s also in the 4.1 presets, so the ambilights pass over. So both generations are affected if the ambilight is strong enough to pass over the background image.
Success! Here is the solution: I put a 1pixel black frame around the artwork and set both the background_over and the background_under to clamp_to_edge, et voila, the ambilight no longer shines through! Let’s close this, that’s actually perfect. So my previous post from some 40 min ago is irrelevant.
I was bored and played around with the clean scanline shader on my phone. These are meant to be for my retroid pocket 3+. That’s why the image is so small (1334x750). I must say I’m really impressed how good it looks on a 720p screen. The jpg’s lost a little bit of quality due compression.
Ahah, I was just trying the same thing on a boring sunday afternoon 
Did you forced tate mode? smart!
my phone can’t handle this, even after switching flick_and_noise_pass frpm 2x to 1x scale
Tate mode could work too with vertical scanlines I didn’t tried that yet. I was using the clean scanlines classic take shader and changed the scanline phosphor gamma from 4.2 to 2.2. This makes the scanlines look normal on a 720p screen. I also added some gamma to compensate for the brigthess. I also added a phosphor grid with mask type 4 and fxaa.
I think it looks quite good
“Good (?)” news Xbox / Direct3D users:
I’ve read comments about koko-aio not working on Xbox series, so i tested it on a Windows OS with d3d output driver and indeed there were problems (crash on d3d12, heavy flickering on d3d11).
The good news is that I tracked down what the issue is.
The bad ones are:
If you absolutely need to use one of those output drivers, then you need to use the development code (see first post), open the file located in config\config-static.inc with a text editor:
Turn the line:
// #define D3D_WORKAROUND
into:
#define D3D_WORKAROUND
You just strip the leading “//” and save it; then reload the shader, and it should work.
I also opened a bug report here, if you want to track the issue:
From this:
Through this:
To this:
Because of this:
If screenshots do not speak for themselves, it seems we won a match against Mr.Moire!
The trick is to stagger odd/even phosphors by a small margin till the beloved moire pattern fades away.
Pro: It is not yet clear to me how the new parameter “Anti-moiree sweet spot” is related (but it is and I’ve yet something to try) to the output size and the core resolution, but a value of 0.17 seems to do a damn good job at 1080p for low resolution content.
Cons: Since it staggers phosphors, it is not compatible with y deconvergence settings, which is auto-disabled when using anything > 0.0 for Moire mitigation.
Innovative. I admit the mask now looks a bit weird staggered like that but the end result is worth it from what I see.
Indeed, if you (like me!) like to watch the phosphors one by one from near, it could be disappointing, but 1 meter away it does look like a normal deconvergence to me.
First thing I did was zoom in to see how the mask looked lol. But yeah from gaming distances it definitely works especially for a game like OutRun.
This is really awesome!
Just for fun. Monitor-Screen-hmask-aperturegrille shader on a t618 Android device. 1336x752 resolution. I just enabled fake integer scanlines. I still can’t belive that it runs smooth
Another one. Sharp-fxaa-core-slotmask This one lags a little
Since the lastet in develompent shader pack fake integer scanlines don’t work for me anymore. If I active them it gives me a blurry picture with vertical scanlines.
Thanks! It is nice to have such speedy bug reports, so it is easier to spot where the regression is.
Lately i was trying to squeeze the remaining squeezable performance juice, but something went wrong.
Will fix by monday or before.
-edit-
@ComfyTsu: it was a matter of minutes, just a stupid typo, fixed.
Are you telling me, you were able to eliminate moire pattern from fill up screen shaders at 1080p?
There are tons of shaders that are moire free at 1080p; it is “just” a matter of not using “too much ®” scanlines gaps.
What I’ve discovered is that when the image is whole odd/even staggered, you can use deeper/darker gaps because you’re effectively getting rid of straight lines.
The idea seems to work, but it is not perfect nor fully explored; by now, if you see moire, you can play with that parameter and hopefully mitigate the moire.
nice, so which is the name of this preset?
also for your presets i’d like to know if there is a way to configurate mask layout to be for bgr
There is no such preset (yet)
koko-aio supports RBG with preset mask 5.
This, with aperturegrille setups gives you BGR by repeating itself:
-RBGRBGRBGRBGRBG-
However, this is not enough for slotmasks when they are staggered/splitted in the wrong position: RBG|RBG|RBG, but you can draft whatever layout you want anyway (if you read the docs-ng.md text file, it is all explained).
I just updated overmask presets in the main repo with moire mitigation; this is with “perceptual ®” slotmask 
Monitor-Screen_Hmask-Core_SlotMask-Overmask.slangp
While this is without slotmask:
Monitor-Screen_Hmask-Aperturegrille-Overmask.slangp
koko-aio NG-1.4 has been released and submitted for merging into retroarch official repo, should be available soon via online updater.
Note on updating
Before running the shader online updater, I suggest you to delete bezel/koko-aio.
This ensures that no stale shaders/presets/files will pollute the folder.
Changelog:
Misc
Presets
Parameters
Fixes/Changes
Performance
