madVR - high quality video renderer (GPU assisted)

[Akeno]

Originally Posted by huhn

do you get this problem only with madVR?

and why do people still use CUVID? DXVA copyback is just better.

Are we referring to the H/W acceleration in LAV Video? Wouldn't CUVID be the most efficient if the user has an NVIDIA card?

[mrcorbo]

Originally Posted by Akeno

Are we referring to the H/W acceleration in LAV Video? Wouldn't CUVID be the most efficient if the user has an NVIDIA card?

Nope.

[Akeno]

Originally Posted by mrcorbo

Nope.

Well then. This brings up other questions I've never thought about: LAV video forces you to dither videos. How does that work with madVR is also dithering videos? Does the video go through two dithering processes?

[ryrynz]

Originally Posted by Akeno

Well then. This brings up other questions I've never thought about: LAV video forces you to dither videos. How does that work with madVR is also dithering videos? Does the video go through two dithering processes?

LAV defers to madVR for dithering by default, also LAV will only dither when required say when converting from 10-bit to 8-bit.

[Warner306]

@madshi

Can you take a look at rendering stats with Kodi DSPlayer? Upscaled videos (720p -> 1080p) are reporting rendering stats of 1-2ms. This is with v0.88.20.

For anyone interested in giving Kodi DSPlayer a try, a build of Kodi v15 Final is now available.

DSPlayer Download Link: http://forum.kodi.tv/showthread.php?tid=223175

Detailed set-up guide in my signature.

[Telion]

Also reporting unrealistically low rendering stats in 88.20.

[Akeno]

I've made a short comparison along the lines of madshi's post comparing Super-XBR and NNEDI3 in regards to SuperRes.
Source chosen specifically for the details in the eyes. These get extremely soft when upscaling.

GroundTruth -|- 50% downscaled


Without Refinement
NNEDI3 -|- Super-XBR

I'm using s-xbr 100 for these comparisons. NNEDI looks a better than s-xbr. You can see that the lines are a bit thicker especially in her eyes and there's small ringing around her chin and on the flash cards she's holding. S-xbr is a bit sharper than NNEDI in this example but NNEDI is closer to the groundtruth here.

SuperRes: Strength 2 / Radius 0.66
NNEDI3 -|- Super-XBR

With SuperRes, the images look nearly identical. NNEDI benefits a tiny bit from SuperRes. The image is slightly sharpened at the cost of tiny aliasing all around. You can see it on the lines of the flashcards. Super-XBR benefits a lot more from SuperRes as the thickness around the lines is significantly reduced but the ringing is still present.

Conclusions
So here's SuperRes on cartoon and anime content. No horrible directional artifacts to be found unlike the trees in the castle examples. All in all, NNEDI3 and Super-XBR produce nearly identical results when combined with SuperRes. NNEDI3 is still superior though as Super-XBR does produce more artifacts (look at the lower edge where the glass meets the door frame next to her shoulder and on the edges of her uniform) and ringing. If the image could be sharpened just a teeny bit more, we'd be extremely close to the groundtruth. Maybe a change in finesharp's algorithm to decrease artifacts or a denoise filter with finesharp could do the trick.

I'm also going to take back what I said about adaptive sharpen as an image enhancer. It completely ruins the soft glow she has on her face and makes the image look processed.

[har3inger]

Akeno: Have you tried using higher sharpness (125, 150) Super-XBR to try and match the sharpness you get from nnedi3?

These comparisons are very interesting. It seems like perhaps SuperRes itself is the primary determinant of final image quality. Might be worthwhile to do some tests of jincAR or even bicubic75AR to allow more headroom for more SR passes.

[pureocean]

Dear @madshi,

Firstly, thank you very much for this renderer. It's very useful. Especially on NT6 and NT10. Because of EVR can't support properly seeking, jumping in videos (it's happening shutters when seeking). Why important to me? I can't reading subtitles fast. Therefore I press often back (left) key in media players.

I have black screen problem. My GPU is Radeon HD 3200. I use Potplayer (last version). Catalyst display driver is last version. A sample video. When in this video jump to 07, 08, 09, 10 and 18, 19, 20 seconds, it's displaying black screen for an instant. Which settings can fix this problem? I don't know using how "madVR debug.ax".

PS: I'm sorry, my bad English.

[Braum]

I did a short comparison of upscaling refinement options, mainly for myself, needed to know what was best for anime content.

Source : 1280x720
Monitor : 1680x1050

Chroma upscaling : Super-xbr 75
Image downscaling : Lanczos 4 AR LL
Image doubling : Super-xbr 75 (double luma/chroma)
Image upscaling : Jinc AR

Zoom 400%

Original

SuperRes Strengh 2 Radius 0.65 (the higher the strengh, the worst the result)

Adaptive Sharpen 0.3

Adaptive Sharpen 0.1

Finesharp 1.0

So SuperRes mostly damage the original content, adaptive sharpen 0.3 is way too much, so I'll go for Finesharp 1.0/1.5 or Adaptive Sharpen 0.1/0.2.

Finesharp seems to be better for movie content (bluray, high quality source) and adaptive sharpen for anime content (relatively new animes), at least that's my preferences.

edit : I decided to go with adaptive sharpen 0.2

[6ari8]

Originally Posted by madshi

That sounds quite interesting. I don't really have an explanation for that, unfortunately. Looking at my code, I seem to be doing everything correctly. Also on my PC I can't reproduce these issues. As a test I've setup GPU queues to 16 and the present queue to 8, and playback is mooth with D3D11 in both windowed and FSE mode, regardless of whether the OSD is on or off.

It seems there are some odd problems out there atm, with v0.88.17+, but it seems to affect only few users, and every of them seems to have different problems, and I can't reproduce any of them. So it's really hard for me to do anything about it. Of course I could simply revert all changes I did, but then we would lose some important improvements that several media player devs have been wishing for and been quite happy to see introduced in v0.88.17 (like low latency OSD, smooth and low GPU power paused mode rendering etc).

I believe the stuttering issue started with v0.88.16b.

Maybe you can ask users who're having similar issues to try and see if they don't happen in debug mode. If the debug mode has the same code as release mode then it really is strange.

Maybe I can try and help you reproduce it.
*enable FSE mode and use D3D11
*Set both CPU queue size and GPU queue size to 10
*Set "how many video frames shall be presented in advance:" in exclusive mode settings to 2
*All trade quality for performance options are disabled.
*Set your monitor/TV to 23Hz
*Download this file: http://usersfiles.com/ltuz45pmu6wz and play it while you have the OSD up.


Also, regarding v0.88.20, the issue is not fixed unfortunately.
It also introduced this issue

Originally Posted by Warner306

@madshi

Can you take a look at rendering stats with Kodi DSPlayer? Upscaled videos (720p -> 1080p) are reporting rendering stats of 1-2ms. This is with v0.88.20.

For anyone interested in giving Kodi DSPlayer a try, a build of Kodi v15 Final is now available.

DSPlayer Download Link: http://forum.kodi.tv/showthread.php?tid=223175

Detailed set-up guide in my signature.

Originally Posted by Telion

Also reporting unrealistically low rendering stats in 88.20.

[ryrynz]

With regards to what Nev said over on the LAV thread

Originally Posted by nevcairiel

With madVR, using Copy-Back may give you slightly better quality, as madVR has a few problems getting access to the unprocessed image otherwise.

Can you give us any firm info on this?

[aufkrawall]

Originally Posted by madshi

Where's the ground truth for this image? Do you *want* our discussion to not be objective? The problem with not having a ground truth is that we can discuss for hours how the upscaled image should look like without ever coming to an agreement.

I don't have any for this example, then let's skip it.

Originally Posted by madshi

Maybe your very tiny original image was downscaled with a very soft algorithm?

As I said, the foreground looks totally clear to me in its native resolution.
Why is this not a valid point, according to you?

Originally Posted by madshi

But artificially sharpening video is not something that should ever be enabled by default in madVR.

And you are missing my point. I never claimed that I would want to achieve artifical sharpness.
Is your display still sub-1080p? I really doubt it makes sense to judge about sharpness if you can count every single pixel.
When upscaling to WQHD, the blur becomes very intrusive when watching from the near. This is not the case when watching 1080p on the same display without scaling, no matter how close I get. So I must be bothered by the blur that emerges with scaling.

Sorry, two years ago I was still on 1280x1024 and felt perfectly happy with it.
Funny thing is that now for me everything looks blurry on it, there is no such thing as "natural sharpness". Instead, it's extremely easy to see any kind of artifacts though.

Probably someone with a 4k display <30" will even judge more in my way since WQHD is still a joke compared to mobile devices.

Originally Posted by madshi

On the other hand, faithfully upscaling sources, by producing images which are as near to the ground truth as possible, that is something that would make a great default setting in madVR.

Of course this should be the goal, but then you'd also need to make NNEDI3 64 the default settings because everything else introduces tons of artifacts.

Originally Posted by madshi

But please have a look at all the lines in these images. Take a step back from your monitor and compare the SuperRes image with the original frame, and with the AdaptiveSharpen image. Here's an extract using super-xbr for upscaling:

I already stated that NNEDI3 produces softer lines than super-xbr (at least super-xbr 100, difference is extreme). I wouldn't recommend any sharpening at all when upscaling without NNEDI3 because nobody needs sharpen of obvious artifacts.
I don't think it makes sense to discuss about your examples without knowing the exact super-xbr and AS settings.
Is it super-xbr 75? Then you have a point. Else not, there is not a real difference in line thickness when comparing 720p -> WQHD NNEDI3 64 + 0.2 AS UR vs. super-xbr 100 no UR.

[daert]

I have a problem with MPC-BE and madvr. If I set "treat 25p as 24p" in madvr, display mode won't change to 24p when I open a 25p video: it's stuck at 60p. If I use MPC-HC the issue doesn't appear. I'm using MPC-BE 1.4.5.579 and madvr 0.88.20

[Akeno]

Originally Posted by daert

I have a problem with MPC-BE and madvr. If I set "treat 25p as 24p" in madvr, display mode won't change to 24p when I open a 25p video: it's stuck at 60p. If I use MPC-HC the issue doesn't appear. I'm using MPC-BE 1.4.5.579 and madvr 0.88.20

That option requires ReClock to speed down the video first. Make sure the PAL speed down option is checked in the ReClock config.

[daert]

Indeed. PAL speed down is already enabled. In fact, everything is fine with MPC-HC

[leeperry]

Originally Posted by 6233638

Here's one of the images that I was testing with which seemed to show aliasing quite well:

• Source
• Downscaled

I'm not really happy with how that looks using any of the sharpening options, or scaling other than NNEDI3.

Homogeneous black outlines really chew NNEDI3's work, it's indeed a strong contender for anime.

Originally Posted by madshi

No, it's got nothing to do with dithering.

I did mean sharpening, I blame the crazy heat

Originally Posted by madshi

Yes, might be an interesting test.

Righty, took me a while to find the ideal picture but I think this one works very nicely:

original untouched 1080p BD screenshot(not captured by me):

downscaled to 960*540 in mVR .15 using CR AR LL and monostatic ED2@8bit:

• in mVR .15:

the untouched 1080p BD capture played in mVR using monostatic ED2@8bit, that's our "ground truth":

sxbr75+SR3@0.41LQ(my favorite)

sxbr75+SR3@0.41HQ

NEDI+SR3@0.41LQ

sxbr75+AS0.5

NNEDI3 for luma+chroma@16 neurons +SR3@0.41LQ

NNEDI3 for luma@32 neurons+J3AR chroma+SR3@0.41LQ

NNEDI3 for luma+chroma@32 neurons

• in mVR .20:

sxbr75+SR1@0.66LQ

sxbr75+SR1@0.66HQ

sxbr75+SR2@0.66HQ

sxbr75+SR3@0.66HQ

All the pics as a big zip file.

I'm literally tired of all that pixel peeping so I might be completely wrong for all I know but I did run rather extensive comparisons on real world 25p material beforehand and this screenshots galore pretty much confirms all my impressions:

-I believe the blur I'm seeing with HQ Vs LQ SR is very obvious and I don't see how HQ could remotely be considered any closer to the original file
-sxbr75 is very nice, I wish one day I could try 70 or 80
-NEDI got superseded by sxbr75 as it would appear
-my favorite one remains sxbr75+SR3@0.41LQ in mVR .15, now that I see those screenshots some very slight AS on top might be a good idea so I will have to give it a try next time
-screenshots are nice and all but the magic SR does on motion blur makes it look far more natural to me(especially 24/25p) so I'm still very keen on sticking to .15 with 3@0.41LQ if that's my only option
I would gladly(and more than likely will) be proven wrong and I'm here to learn so shoot me now if you must, kthx

[Akeno]

leeperry, taking a look at your examples, I'm still under the impression that HQ is closer to the groundtruth than LQ. While LQ gives a nice impression of the shadows underneath the roof tiles, it thickens the lines everywhere else too much. There's also some strange artifacts around the windows of the closer house. The small black chunks seem to be present in every example you post but LQ turns them into definitive lines.

[har3inger]

Originally Posted by ryrynz

With regards to what Nev said over on the LAV thread



Can you give us any firm info on this?

This may be GPU/driver specific. The only way to test for yourself is to take two screenshots of the same frame, one with DXVA2 native, one without, and subtract them from each other in photoshop or something.

On my setup, if you compare DXVA2 native and DXVA2-cb, there are definite differences if you subtract the screenshots from each other that are greater than the difference you get from the different dithering pattern. All of these differences are barely visible, as they are under 5/255 luminosity in their intensity.

I'm on an ATI HD 8870 (R9 M270X) paired to an intel HD4000. Catalyst 15.6 beta

[strangeluck]

Originally Posted by Telion

Also reporting unrealistically low rendering stats in 88.20.

Same here using MPC-HC. Render times are off by a minimum of 20-30ms. I have never had problems with render time display until the most recent round of updates. It seems to give accurate results when playing video on my primary display with the secondary monitor (a 1080p Samsung television) disabled. Playing video on the secondary display always shows render times of 5ms or less which is just entirely incorrect.