version 1.3 was the first one where I implemented a specialized checkerboard detection.
so there is a huge difference especially in those areas of a picture. there v1.5 definitely should outperform v1.2.
I’m actually working on a more generalized concept which isn’t so strict with the colors but is still only making small and fast local decisions.
here is a wip shot of a mask which assigns every pixel a gray value which stands for the probablity that this pixel is part of a dithering pattern:
my idead so far is to propagate this value to the neighbors and mix them with their own (to eleminate false detections).
and in the last step merge only pixels whose value is over a certain threshold.
I’m testing mdapt 1.5 on RA for PS3. It works very well with checkerboards and pseudo transparencies for arcade games, though some transparencies merge with its borders in some cases that looks like an line of sight next-gen effect. The only thing it doesn’t work on PS3 very well is with genesis pseudo transparencies (vertical lines).
Look at these arcade games and tell me if it’s supposed to look like this:
the screens from the arcade games look ok, but the ones from the earthworm jim don’t.
don’t know the game too well but just to confirm that if it’s mdapts flaw or some error on your side, here is a screenshot from lion king (first stage, right at the start) with “hori=true” and “vert=false”:
if it looks the same on your ps3 there is maybe a problem with different pattern detection in the same area.
just gimme some feedback and I’ll look into it.
I think I have a clue about the problem. PS3 doesn’t work with float_framebuffer enabled.
But, I’d like you test to me a xBR multipass I was making for PS3, but as it doesn’t support float_framebuffer, I can’t verify if it works. The output isn’t good in that console. As you can test on PC (I can’t), maybe it could work. It’s just a two pass combination.
Thank you. Those glitches are the same I’m getting on PS3. So I think I have to do some tests myself and if I can’t find the answer, then I’ll post another test in the xBR thread later.
Kkkk. Actually, I disabled vertical and horizontal line detection. They’re responsible for pseudo transparencies in genesis games. (see sonic waterfall).
Not crazy about xbr, but loving the dithering blending.
Is there going to be an attempt at tackling the kind of full screen checkerboard dithering found in many ps1 or n64 games? Silent Hill specifically is what I’m thinking about. I always recommend people play that on real hardware on a CRT TV, since it’s very difficult to set that game up right via emulation. The dithering is one aspect. Usually it’s turned off, which lightens the colors in the game, making them paler, or it’s unblended.
Also be aware there’s a page on dithering on the Emulation Wiki, with a sub section for this very shader. Feel free to review or edit it
For us dummies , could you please include a link for this UPDATED filter in a compiled format? (on my mac I just can’t get github to download properly)
In any case, does this play better with text? The original MDAPT on this forum works well with dithering but creates artifacts with text, etc.
Also, this new combo doesn’t treat pseudo transparencies (sonic waterfall), but is much better at dithering than the old one. Besides, it is better at dealing with fonts.
What image should be added to the Dithering wiki page? I’d like something that shows off the latest version. No xbr, since that’s another shader, and I just want to show off the dithering shader itself. I was thinking of using an image from the thread, but those are probably older versions.
I didn’t take it. I don’t know what emulator or settings were used.
Wow, you fixed the dithering adapt! Thank you SO much for this!! In many ways, this was the last missing piece of the puzzle for me.
Thank you…
Thank you…
Thank you…
It’s created a new dilemma for me though… :-/ haha…
I really REALLY love your original BETA 5xBRdilation filter (since it gets things ‘just’ right on most games in terms of sharpness…) In other words, it’s JUST sharp enough to be perfect, something I’ve always wanted. That’s why I like it so much. It’s very balanced…especially on older HDTVs like mine that seem to be a bit blurry than i’d like.
However I’m now in love with your new four pass dither filter…so…
The problem I have when trying to COMBINE the four pass dither WITH the 5xBRdilation is that it’s a touch too slow…still very playable, but just slow enough to be annoying for an OCD guy like me…lol…
Yes, I did try the 4-pass-XbrC filter preset and while it looks great it seems to lack just that extra little bit of sharpness that makes the beta 5xBRdilation so perfect.
I know you made the separate dilation one pass filter to enhance the other XBR but for whatever reason it just doesn’t look as good as that beta 5xBRdilation that you first put on there IMHO. Not sure what you did there but beta or not it’s close to perfect as perfect can be IMHO. Maybe it’s a happy accident but it’s worth another look, it really gets things right…
That said, If you have any plans to go back and rework 5xBRdilation so it’s not as slow, that would be awesome!!! I’m not sure how this all works so I’m not sure if that’s even possible to optimize it so it runs faster. I dont know how to code so I’m not sure what went into all that but it really does the trick. Kudos to you for that.
Like I said, to be able to add ‘dithering’ to that would be a match made in heaven (which creates a heavenly look, despite the not so pleasant speed constraint).
So…for me the perfect ideal filter COMBO at this point would be:
-mdapt dither pass 1
NEAREST
X1
-mdapt dither pass 2
NEAREST
X1
-mdapt dither pass 3
NEAREST
X1
-mdapt dither pass 4
NEAREST
X1
-5xBRdilation
NEAREST
NOTHING
This is what I’m using now for most things…
In fact, as far as the cores go, it works great with GBAnext, okay with T16 (and CDs), and a little slow with Arcade and Genesis (playable but just slow enough to be a little annoying).
Again, this is just my 2 cents and I’m no programmer so I’m coming at all this as strictly a novice. But in a perfect world (for me, at least), fixing that old beta 5xBRdilation filter to run ‘a touch faster’ is the Holy Grail for me and Im sure alot of retro enthusiasts that are a fan of your work. So I hope it’s okay that I shared my opinion about all this.
You’ve done some remarkable work and I’m surprised the guys at Sony haven’t called you yet asking what your secret recipe is for their lame attempts to add HD filters to their re-releases when so many of them are bad (your’s are superior).
If you take donations, you should post a link since it’s a shame you are not getting paid for this. Your work is truly remarkable and breathes new life into these old games.
Well, thanx for everything. Think of this post as a glorified wish list for his obsessed OCD retro ethusiast here looking for the ‘perfect look’ on his game. I feel like I’m a couple of frame rates away (in terms of speed) to getting where I want to be in all of this and I’m grateful to you for your hardwork. Know that it doesn’t go unnoticed.
P.S. Thanks for also showing me how to download all that from the repo. Much appreciated. Ciao.
Sorry to point this, but the credit must be given to the right person.
mdapt is an original idea from Spookyfox, not me. As I like it too much, I try to optimize it to work well on most RA platforms.
Yes, I know. I have some ideas to optimize it.
Can you take two game screens showing why you think xBRdilation is better than xbr separated from dilation.cg? I really don’t know the reason as the multipass version do the ideal dilation.
Ok, download these two files (a shader and a cgp) and put insided xbr-lv2-multipass folder in your Retroarch. Load this cgp and see if the speed is good and sharpness.
, could you please include a link for this UPDATED filter in a compiled format? (on my mac I just can’t get github to download properly)
EDIT: I also wrote 
Ciao.