madVR - high quality video renderer (GPU assisted)

[leeperry]

OK, so both 10 & 12 bit 4:2:2 seem possible in HDMI 2.0b but do the RX480 drivers provide an option to do that? FWIR only nvidia allow this kind of kludge in RGB?

Actually that link seems to imply that it would even be possible with HDMI 1.4, it's all pretty dang confusing and I haven't seen any option to output 4:2:2 RGB in the 13.12 AMD drivers.

[Asmodian]

No one offers 4:2:2 in RGB, that doesn't make any sense. This is all YCbCr 4:2:2, which is fairly common but your display needs to report support for it.

[kitame]

Originally Posted by cork_OS

Currently NGU pix v3 exists only in form of very high quality settings.

wait, then what am i getting from V3 low? original NGU? or NGU PIX V1?

well aside from that, i find my current settings acceptable for a balanced point, except it heats up my HD7950 by quite a bit.

[ryrynz]

Standard ol' NGU. Waiting hard on this proper NGU pix release in a week or two.

[sauma144]

Originally Posted by ryrynz

Waiting hard on this proper NGU pix release in a week or two.

Why not today?

[cork_OS]

Originally Posted by sauma144

Why not today?

It's too good to be true.

[madshi]

Originally Posted by 70MM

Sorry madshi I forgot I have a macro named that but its actually activating your (Zoom Control) if there are big black bars: dropdown box ....reduce bar size by 25%

This is what I use labelled Zoom 5. This removes the black bars on older 2.40, 2.55 titles like Ben Hur, but its removing too much. Im wondering if we could have some smaller steps up to 25% please, maybe 5%, 10%, 15%, 20%, then 25%.

I guess that should be possible, I'll add it to my to do list.

Originally Posted by leeperry

Is there any way to check that 10bit got enabled

The only reliable way I know is if your receiver or TV report the incoming bitdepth. If madVR sends 10bit to the GPU driver, the GPU driver could still ouput it as dithered or undithered 8bit, 10bit, 12bit or 16bit. I've no control over that, I can't even ask the GPU which bitdepth it actually outputs.

Originally Posted by ryrynz

Nah, he just says 8.1 works well, but then so does 10.

I'm actually saying Windows 8.1 is the best media playback OS atm. I see zero improvements in Windows 10 for media playback, and most users who have problems have them with Windows 10 and not with Windows 8.1.

There may be HDR APIs for Windows 10 soon, but on the other hand, I've heard about other planned changes which are totally terrible. So I keep recommending to use Windows 8.1 for HTPCs.

Originally Posted by Tarzan

This Wednesday I am finally getting a projector with lens memory function (jvc x35) and I am now making different profiles according to aspect ratios. However I wonder whether madvr after playback automatically changes the lens memory back to the whatever it was before playback.

That's currently not supported. E.g. imagine you have a 16:9 and a scope lens memory, and your default setting is scope. Now imagine you play 3 scope movies in a row (e.g. LOTR trilogy). If you tell madVR to restore the lens memory after playback, what you'd get is madVR switching to scope at the start of the first movie, and then between movies 1 and 2, and between movies 2 and 3, madVR would first restore the 16:9 lens memory, and then switch back to scope a couple seconds later. I don't think that's a good idea.

I'm using an X35, too, btw.

Originally Posted by XMonarchY

development seems to have slowed down

Huh??

Originally Posted by wasper

What is video mode exactly? For me, video mode, is just not film mode. So maybe, it does something more that could bring me other benefits. I searched forums and wasn't able to find a clear answer.

"Video mode" is not a fixed term. But I suppose some people using this term might mean that the deinterlacer is turned into a mode which it treats all content as natively interlaced video content.

Originally Posted by wasper

Ok, I'm trying to use film mode as much as possible. However, I have some videos that seems to be 3:2 pulldown but not perfectly. Playing them in film mode will bring a lot of cadence break and dropped frame. On the other side, if I'm deintrelacing the 29.97 video to play it at 59.9, it won't play well as there is a lot of pulldown judder like you said.

How many is "a lot"? Do you see the cadence breaks and dropped frames? Or is it just the meaningless statistic in the OSD that bothers you?

Originally Posted by wasper

So, why my blu-ray is able to play this DVD correctly and I'm not.

Does the blu-ray player able to play it smoothly, without pulldown judder, and without any (visible) cadence breaks and frame drops?

Originally Posted by Oguignant

Does anyone know how to create a 3d lut file? Any web that explains it?

I use Calman 5 software, but I do not have good results.

The big question is: Is this likely a problem with madVR? Or a problem with Calman 5? Or a problem with your HTPC configuration? This thread is the right one if you suspect a bug in madVR. In the other two cases you'll likely get better support by asking on the Calman support forum. Most of us here are not really calibration experts. So if Calman is for some reason not creating a good 3dlut, you'd better ask them about it. Of course if the cause of the trouble is a bug in madVR, you'd be better served here in this thread, but it's really hard to say from your description.

Originally Posted by IggySLO

Is there a way to make the seek bar larger in exclusive mode?

It's planned for a future version, but not too soon.

Originally Posted by TheShadowRunner

Hi madshi, if you have a little time please don't forget the test build with "* last video frame is now remembered for 2 seconds when stopping graph" disabled (for the DVD on XP bug).

Noted.

Originally Posted by Razoola

Is there any reason why with the following in madvr's monitors display properties, it will select 2160p30 for 25hz material?

2160p23, 2160p24, 2160p25, 2160p29, 2160p30

See here:

https://forum.doom9.org/showpost.php...ostcount=42797

Originally Posted by sarah180

Hey @madshi. I would like to request that you restore the ability to do nearest-neighbor upscaling. It seems to have been removed in recent versions.

I have an actual use case for this, and am sticking with an older version as a result. Although nearest-neighbor is not generally a desirable algorithm for most realistic images, it can be desirable when displaying media containing pixel artwork. It is also quite useful as a means to demonstrate to people the differences between different interpolation kernels on real media, something I have occasion to do in an educational capacity.

My suspicion is that was removed because you imagine that nobody uses it and you could save real estate. I can testify that I use it regularly (though not for photographic content). Perhaps you could combine nearest-neighbor and bilinear upscaling under a shared radio button and name it something like "Simple," much as you folded C-R and M-N under the "Cubic" banner. (Another idea would be to combine NN, bilinear and DXVA2 under a "Native" or "GPU" grouping.)

Thank you for all your hard work on publishing madVR for free: it is a fantastic offering. I know you can't fulfill the requests of everybody, but I hope you will at least consider restoring this functionality. It would brighten my day.

I only ever included it for "educational capacity", as you say. I don't think it's useful for anything else, even for pixart, unless you absolutely need to check every single pixel separately in pixart, but why you need that? Check out this one:

http://screenshotcomparison.com/comparison/202526

Why would you ever want nearest neighbor, if you can get the quality shown in that screenshot comparison, when using madVR?

Originally Posted by LexSfX

I guess you don't understand. Seeing the individual pixels is important in watching retro speed runs; especially tool-assisted speed runs, as pixel-perfect play is common there. I don't want a blurry image or distorted-shaped pixels when upscaling such videos. Also, I agree that nearest-neighbor is relatively bad, as my post explained. The algorithm I linked to in my post is a superior algorithm to nearest neighbor, as it hides the misshapen-pixels problem nearest neighbor has without eliminating its crystal clarity.

I'm confused. You're linking to some algorithms which distort the pixels but you say you don't want distored-shaped pixels?

Why exactly do you need pixels to be exactly the way nearest neighbor produces them? It seems weird to me. I can understand that you don't want blurry, but you don't need to have blurry.

Originally Posted by LexSfX

Nearest neighbor is the only truly sharp algorithm provided by madVR for low-res pixel art video (256×224, for example)

Huh?

http://screenshotcomparison.com/comparison/202526

To my eyes nearest neighbor is not sharper in that screenshot comparison. It's just dramatically more aliased.

[madshi]

madVR v0.91.6 released

http://madshi.net/madVR.zip

* added NGU Anti-Alias algorithm, intended as a potential NNEDI3 replacement
* completely redesigned image upscaling settings page once more
* downscaling after doubling now uses relaxed AR (instead of strict)
* increased AdaptiveSharpen + LumaSharpen upper limits
* fixed: up & downscaling anti-ringing checkbox was sometimes disabled
* fixed: 3D subtitle depth was incorrect for movies with swapped eyes

Hope you guys are happy with the upscaling settings redesign? I'm not fully happy with that there are so many options now, but I suppose defaulting them all to "let madVR decide" should help newbies, so I hope it will be ok for a while.

I've renamed "NGU pixart" to "NGU Anti-Alias" to make sure that there are no misunderstandings. NGU Anti-Alias is not primarily made for "pixart" (as in old computer games where each pixel was carefully drawn by an artist), but it's optimized for heavily aliased photos and videos. It also does happen to work reasonably well for pixart, but that's not the primary target.

NGU Anti-Alias is an improved version of "NGU PixArt v3". Compared to the PixArt v3 test build, it has slightly better anti-aliasing capabilities, while keeping the same (or higher) focus and sharpness. NGU Anti-Alias is now available in all 4 quality levels.

You can see that there will be "NGU Soft", "NGU Standard" and "NGU Sharp". These will be 3 different NGU variants of the original (non-pixart) NGU, which have different sharpness levels built into the very algorithm. The purpose of this change is that we won't need a separate "soften edges" algorithm, anymore, which saves some performance. Furthermore, "soften edges" doesn't work as well for direct quadrupling as it does for doubling twice, which will not be a problem for NGU Soft/Standard.

FWIW, "NGU Soft" and "NGU Standard" are not ready yet, but the settings dialog already mentions them. "NGU Sharp" is currently the same as the original NGU algorithm.

Now I'd like to have your feedback about the following topics:

1) Is the new upscaling settings page fine now?
2) When disregarding speed, how would you rate NGU Anti-Alias Very High vs. NNEDI3 256-taps, for image upscaling?
3) When taking speed into account, how would you rate NGU Anti-Alias vs NNEDI3 at the various quality levels, for image upscaling?
4) Same as 2) and 3), but for chroma upscaling.

[amayra]

so we have more options for upscaling this is good news that why i love madVR
thanks for you hard work i well do some test and comeback with feedback

[Neo-XP]

Originally Posted by madshi

1) Is the new upscaling settings page fine now?

It looks complex when you first see it, but then it is ok (for me).
One thing you can improve : when you choose "disabled" for "<-- luma quadrupling", the list box for "<-- quadrupling" should be greyed out.

I still can't choose "Lanczos3 AR" for chroma doubling. I'm stuck with "Bicubic60 AR" or "NGU Sharp (low)"
Can you add for chroma doubling an option "same as upscaling algo", if you don't want to change the default algos ?

Originally Posted by madshi

Nothing strange about it. We already discussed earlier that I'm using strict downscaling now, which I might change to relaxed in the next official build.

"NGU-Sharp (high)" is still not as sharp as "NGU-med" in v0.91.1. Is that normal ?

[madshi]

Originally Posted by Neo-XP

I still can't choose "Lanczos3 AR" for chroma doubling.

I've yet to see a screenshot comparison where Lanczos 3 AR is visibly better than Bicubic60 AR for chroma doubling.

Originally Posted by Neo-XP

"NGU-Sharp (high)" is still not as sharp as "NGU-med" in v0.91.1. Is that normal ?

Using exactly the same downscaling algorithm? Which image are you testing with? I need it in original (unscaled) size.

[tobindac]

Originally Posted by madshi

1) Is the new upscaling settings page fine now?

It seems fine for power users. You might want to create an alternative/default/first impressions interface for very novice users (with basic options like "Sharp Upscalers (<type1>)", "Sharp upscalers (<type2>, "Soft Upscalers" etc). madVR is getting popular and people start using it en masse and they are not going to like having to read much about it.
Heck even other programs start copying it that they don't know much what they're doing *cough*mpv*cough*.

[Neo-XP]

Originally Posted by madshi

I've yet to see a screenshot comparison where Lanczos 3 AR is visibly better than Bicubic60 AR for chroma doubling.

It is slightly sharper and the closest to NGU (NGU is way sharper, but demands too much resources). Also, Bicubic60 AR and Lanczos3 AR are identical in term of performances.
See https://forum.doom9.org/showthread.p...83#post1790283 for original image
and https://forum.doom9.org/showthread.p...72#post1790272 for a zoomed comparison.

With Cr rendered as Y (by huhn) :
Bicubic60 AR : https://abload.de/img/1dup4r.png
Lanczos3 AR : https://abload.de/img/24arvq.png

I am sure that you can see it better with some well-picked patterns...

Originally Posted by madshi

Using exactly the same downscaling algorithm? Which image are you testing with? I need it in original (unscaled) size.

There is no downscaling algorithm used, only upscaling with NGU. All the other settings are the same.

original : http://www.mediafire.com/file/ce28j9...h/original.png
v0.91.1 : http://www.mediafire.com/file/gv5wl4...wf/v0.91.1.png
v0.91.6 : http://www.mediafire.com/file/2m6n7b...5h/v0.91.6.png

Once again, you need to zoom quite a lot to see the difference, but there should not be one.

[aufkrawall]

Great job on NGU AA, to my eyes it looks better than NNEDI3.
Good thing is that it apparently can also be combined with deringing filter without nasty side effects. I think that's the way you want to watch bad quality sources, also a good fit for chroma.

[tobindac]

PS. For the interface, big tooltips after hovering over algorithms/methods/settings, that explain what each of the setting/algorithm/method is might help. It could be kinda cryptic but perhaps helpful and not distracting if it pops up only after 2seconds. A half-decent solution I had devised was to have a "help" indicator next to each option that would automatically hover a very decriptive tooltip about it which makes it more explicit that that documentation is available so the work for it doesn't go to waste.

[leeperry]

Originally Posted by madshi

Is the new upscaling settings page fine now?

Thanks a lot for the new build but could you please bring back the "use image downscaling" option for downscaling doubling? My fav combo on many noisy videos is sxbr75+SR2 AR LL+SSIM 2D 100% LL without AR. Everything else looks great, only this option for OCD'ed video nerds and it'll be fine I guess ^^

[huhn]

i get image corruption with NGU sharp very high X2 -> NGU sharp very high X2.
or with other words when i use NGU sharp 2 times.

i kind of miss "use image downscaling" under image upscaling for the downscaling algorithm.

but downscaling got a lot less important for me now that i can control the supersampling again.

EDIT: for me NGU AA needs a "lot more" performance than NGU pix art 3. this is working as intended?
i can't use it in real time anymore.

[Backflash]

1) Is the new upscaling settings page fine now?
Yes. In my opinion it's very intuitive now.

Is NGU Sharp sharper than 0.91.5? Looks that way. I love that, i'd crank it up even more like one of the first versions though.
Can't comment on pixart changes, I really do not understand them.

[madshi]

Originally Posted by tobindac

It seems fine for power users. You might want to create an alternative/default/first impressions interface for very novice users

Originally Posted by tobindac

PS. For the interface, big tooltips after hovering over algorithms/methods/settings, that explain what each of the setting/algorithm/method is might help.

I've already said this many times before: There will be a proper documentation, and a redesigned settings dialog for madVR v1.0. Until then I try to keep the changes to the settings dialog as small as possible. So any dramatic changes will have to wait.

Originally Posted by Neo-XP

It is slightly sharper and the closest to NGU (NGU is way sharper, but demands too much resources). Also, Bicubic60 AR and Lanczos3 AR are identical in term of performances.
See https://forum.doom9.org/showthread.p...83#post1790283 for original image
and https://forum.doom9.org/showthread.p...72#post1790272 for a zoomed comparison.

Your comparison shows a very very small difference, and you've zoomed the image up to like 1000% to make it more easily visible. I don't find that very convincing, to be honest. How does the comparison look like when upscaling just 2x instead of 10x?

The way the chroma doubling setting is designed atm limits the number of algos I can offer. If I replace Bicubic60 AR with Lanczos3 AR, performance *will* go down for some users with older GPUs. It might not be a big speed difference, but it will be there, and it might be the difference between smooth and stuttering playback. So I don't like the idea to replace Bicubic60 with Lanczos3 at all.

Of course it's always possible to allow both options, but doing so in an intuitive way is difficult. E.g. currently "normal" quality is Bicubic60 AR. So how would I name Lanczos3 AR? Would that be "normal + 0.01% quality"? I don't think it makes sense for the majority of users.

Originally Posted by Neo-XP

There is no downscaling algorithm used, only upscaling with NGU. All the other settings are the same.

original : http://www.mediafire.com/file/ce28j9...h/original.png
v0.91.1 : http://www.mediafire.com/file/gv5wl4...wf/v0.91.1.png
v0.91.6 : http://www.mediafire.com/file/2m6n7b...5h/v0.91.6.png

Once again, you need to zoom quite a lot to see the difference, but there should not be one.

Looking at my source code, the NGU algo was tweaked between v0.91.1 and v0.91.6. The quality presets "medium" and "very high" seem to be identical, but the quality preset "high" was modified slightly. I don't remember the exact reason, probably I found some artifacts somewhere and saw the need to improve "high". Sometimes improving quality in one aspect costs a little bit quality somewhere else. Looking at your image comparison, I have a hard time seeing any sharpness difference. Ok, the reflection in the right eye is a tiny bit sharper with v0.91.1, but that's really a *very* small difference. Probably when comparing with a different image, it might look better with v0.91.6.

Originally Posted by aufkrawall

Great job on NGU AA, to my eyes it looks better than NNEDI3.
Good thing is that it apparently can also be combined with deringing filter without nasty side effects. I think that's the way you want to watch bad quality sources, also a good fit for chroma.

Great to hear that! I'm a bit surprised, though. I thought that the deringing filter made problems with both NGU and NNEDI3. But now you say it works fine with NGU AA? It's surprising because NGU AA is supposed to behave very similar to NNEDI3. So I'm not sure why NGU AA would work better with deringing than NNEDI3 does. But if that's really the case, of course I won't complain!

Originally Posted by leeperry

could you please bring back the "use image downscaling" option for downscaling doubling?

Originally Posted by huhn

i kind of miss "use image downscaling" under image upscaling for the downscaling algorithm.

Ooops. Removing that option wasn't intentional.

Originally Posted by huhn

i get image corruption with NGU sharp very high X2 -> NGU sharp very high X2.
or with other words when i use NGU sharp 2 times.

Is that with "chroma quality" set to "normal"? I get a weird colored image that way. But with every other chroma quality setting the problem doesn't seem to occur. Can you confirm that finding?

Originally Posted by huhn

EDIT: for me NGU AA needs a "lot more" performance than NGU pix art 3. this is working as intended?
i can't use it in real time anymore.

Then some other part of the configuration must have changed because there should be no difference in NGU AA speed compared to NGU pix art 3.

One thing coming to mind is NGU pixart 3 "high" preset didn't exist. It actually used "medium". Could that be an explanation?

Originally Posted by Backflash

Is NGU Sharp sharper than 0.91.5?

It should be identical. However, if you downscale after NGU Sharp, the final result should be sharper now, because the downscaling algorithm uses a less aggressive anti-ringing filter now.

[leeperry]

Originally Posted by madshi

The only reliable way I know is if your receiver or TV report the incoming bitdepth. If madVR sends 10bit to the GPU driver, the GPU driver could still ouput it as dithered or undithered 8bit, 10bit, 12bit or 16bit. I've no control over that, I can't even ask the GPU which bitdepth it actually outputs.

So at the end of the day, when ppl claimed that 10bit off mVR looked better than 8 for all we know they were gazing at dithering in their graphic card drivers or TV? There's no hard-proof test pattern that can validate how things are being sent? And worst case scenario if it's still entirely subjective, I guess A/B'ing as quickly as possible a 10bit gray ramp might help? Considering how much of a crap shot a potential upgrade to RX480 & NGU + HDMI2.0b UHD TV would be for me, right now my only motive to upgrade to W8.1 would be 10bit 1080p60.