Media Player .NET (MPDN) - D3D HQ GPU Video Renderer [v2.49.0/v1.31.0 27 Dec 2018]

[mecedo]

Can you help me with the following issue? I can't enable Hardware Acceleration in LAV Video in MPDN Player. If I set "DXVA Native" in LAV Video then "Available" is displayed and Active Decoder is "avcodec". In Potplayer with LAV Video which is configured in the same way as in MPDN displayed Active Decoder is "dxva2n" and hardware acceleration works.
Is it possible to debug or log MPDN Player to find out why hardware acceleration isn't applied?
Checked configurations: Intel i3 4130T (HD4400 video) with Win7 x32 and Win 7 x64.
Checked Intel video drivers: 10.18.10.3960 and 15.36.28.4332.
Any help will be appreciated.

[Zachs]

Can you help me with the following issue? I can't enable Hardware Acceleration in LAV Video in MPDN Player. If I set "DXVA Native" in LAV Video then "Available" is displayed and Active Decoder is "avcodec". In Potplayer with LAV Video which is configured in the same way as in MPDN displayed Active Decoder is "dxva2n" and hardware acceleration works.
Is it possible to debug or log MPDN Player to find out why hardware acceleration isn't applied?
Checked configurations: Intel i3 4130T (HD4400 video) with Win7 x32 and Win 7 x64.
Checked Intel video drivers: 10.18.10.3960 and 15.36.28.4332.
Any help will be appreciated.

MPDN doesn't support DXVA native. Other modes are fully supported.

[mecedo]

MPDN doesn't support DXVA native. Other modes are fully supported.

Thanks. Maybe you should mention it in description?
Here is useful explanation: LAV Video Configuration (http://wiki.team-mediaportal.com/1_MEDIAPORTAL_1/0_What's_New/1.3.x/1_UI_Related_Changes/Config_Related_Changes/Codecs_and_Renderer/LAV_Configuration/LAV_Video_Configuration)

And another question. Can MPDN use WASAPI as Audio renderer?

[ryrynz]

And another question. Can MPDN use WASAPI as Audio renderer?

Sanear uses WASAPI, I'd recommend using that, it's included in the installer.

Hopefully a new version of that isn't that far off, Alex has been committing some code recently.

[Zachs]

Quick update.

A new MPDN version will be released soon (later today or tomorrow) along with new extensions with plethora of changes (SuperRes/SuperChromaRes, SSIM etc).
Shiandow will do a quick write up on the changes later.

Someone requested a display gamma correction feature some time ago. It, in the form of a shader file to be used with Image Processor, has also made it into this next release.

Cheers.

[Shiandow]

Hi,

A few comments about the upcoming release.


Super-XBR has been updated by Hylian with an improved anti-ringing algorithm and tweaked parameters.
Image Processing no longer requires shaders to be in the 'ImageProcessingShaders' folder.
SSIM performance has been improved, and a limiter has been added to prevent it from amplifying details too much.
SuperRes / SuperChromaRes performance has improved with numerous tweaks to improve picture quality as well. Most of the tweaks are small but together they might make a significant difference. Also added 'Sinc-Jinc' as the default prescaler, since it is faster and sharper for low(ish) scaling factors (below 2~3), which makes it better for most practical purposes. Also added a fast preset for Super-Xbr that gives a good quality / performance ratio for higher scaling factors.
Bilateral now uses a more reconstructive method for retrieving chroma information from the luma, this method is still somewhat experimental and is prone to artefacts, but in some scenarios it blows away all other algorithms.
Lots of other tweaks and fixes (see the release notes / git repository for more details).


Anyone interested in developing may also be happy to hear that the framework has been improved quite a bit for this new release, also improving the readability of the RenderScripts (which should be convenient for people interested in porting them to other applications). Anyone not (yet) interested in developing will hopefully also see some of the benefits soon, in the form of less bugs and more features.

[foozoor]

Quick update.

A new MPDN version will be released soon (later today or tomorrow) along with new extensions with plethora of changes (SuperRes/SuperChromaRes, SSIM etc).
Shiandow will do a quick write up on the changes later.

Someone requested a display gamma correction feature some time ago. It, in the form of a shader file to be used with Image Processor, has also made it into this next release.

Cheers.

Hi,

A few comments about the upcoming release.


Super-XBR has been updated by Hylian with an improved anti-ringing algorithm and tweaked parameters.
Image Processing no longer requires shaders to be in the 'ImageProcessingShaders' folder.
SSIM performance has been improved, and a limiter has been added to prevent it from amplifying details too much.
SuperRes / SuperChromaRes performance has improved with numerous tweaks to improve picture quality as well. Most of the tweaks are small but together they might make a significant difference. Also added 'Sinc-Jinc' as the default prescaler, since it is faster and sharper for low(ish) scaling factors (below 2~3), which makes it better for most practical purposes. Also added a fast preset for Super-Xbr that gives a good quality / performance ratio for higher scaling factors.
Bilateral now uses a more reconstructive method for retrieving chroma information from the luma, this method is still somewhat experimental and is prone to artefacts, but in some scenarios it blows away all other algorithms.
Lots of other tweaks and fixes (see the release notes / git repository for more details).


Anyone interested in developing may also be happy to hear that the framework has been improved quite a bit for this new release, also improving the readability of the RenderScripts (which should be convenient for people interested in porting them to other applications). Anyone not (yet) interested in developing will hopefully also see some of the benefits soon, in the form of less bugs and more features.

You are amazing guys. ;)

[Zachs]

v2.46.0/v1.26.0 Released.

Please let us know if there's any issues.

Note: Your old audio script group config may throw up an error when you try to open the "Script Group" config dialog. To fix this, delete the %localappdata%\MediaPlayerDotNet\AudioScripts.?? (?? == 32/64/AnyCPU) folder and reconfigure your script chain.

This will be fixed in the next minor release. EDIT: Fixed in v1.26.1 (clears the config automatically instead of failing to bring up the config dialog altogether)

Cheers!

[cascardian]

Thank you for the newest version!

This is what I now get in DebugView on a fresh default config with any AudioScript enabled (I apologise if this is redundant for you):
00000002 5.56374979 [2344] [109346'021,584ms] [INFO] Spent 516ms building/loading extensions
00000003 6.08320141 [2344] Information: Create ComputeContext(806126304) in Thread(1).
00000004 6.08431005 [2344] Information: Create ComputeCommandQueue(634353200) in Thread(1).
00000005 6.69643497 [2344] Mono.Cecil.AssemblyResolutionException: Failed to resolve assembly: 'Mpdn.Core, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null'
00000006 6.69643497 [2344] at Mono.Cecil.BaseAssemblyResolver.Resolve(AssemblyNameReference name, ReaderParameters parameters)
00000007 6.69643497 [2344] at Mono.Cecil.DefaultAssemblyResolver.Resolve(AssemblyNameReference name)
00000008 6.69643497 [2344] at ICSharpCode.Decompiler.Ast.Transforms.IntroduceUsingDeclarations.Run(AstNode compilationUnit)
00000009 6.69643497 [2344] at ICSharpCode.Decompiler.Ast.Transforms.TransformationPipeline.RunTransformationsUntil(AstNode node, Predicate`1 abortCondition, DecompilerContext context)
00000010 6.69643497 [2344] at Cudafy.Translator.CUDALanguage.DecompileType(TypeDefinition type, ITextOutput output, DecompilationOptions options)
00000011 6.69643497 [2344] at Cudafy.Translator.CudafyTranslator.DoCudafy(CudafyModule cm, Type[] types)
00000012 6.69643497 [2344] at Cudafy.Translator.CudafyTranslator.Cudafy(IEnumerable`1 props, Type[] types)
00000013 6.69643497 [2344] at Cudafy.Translator.CudafyTranslator.Cudafy(eArchitecture arch, Type[] types)
00000014 6.69643497 [2344] at Mpdn.Extensions.Framework.AudioHelpers.GetCudafyModule(Type[] types)
00000015 6.69643497 [2344] at Mpdn.Extensions.Framework.AudioHelpers.LoadAudioKernel(GPGPU gpu, Type type, Type[] types)
00000016 6.69643497 [2344] at Mpdn.Extensions.Framework.AudioChain.AudioProc.Initialize()
00000017 6.69762754 [2344] Error in AudioChainScript:
00000018 6.69762754 [2344]
00000019 6.69762754 [2344] Object reference not set to an instance of an object.
00000020 6.69762754 [2344]
00000021 6.69762754 [2344] ~
00000022 6.69762754 [2344] Stack Trace:
00000023 6.69762754 [2344] at Mpdn.Extensions.Framework.Chain.FilterChainScript`2.Execute()
00000024 6.70564222 [2344] Error in AudioChainScript:
00000025 6.70564222 [2344]
00000026 6.70564222 [2344] Object reference not set to an instance of an object.
00000027 6.70564222 [2344]
00000028 6.70564222 [2344] ~
00000029 6.70564222 [2344] Stack Trace:
00000030 6.70564222 [2344] at Mpdn.Extensions.Framework.Chain.FilterChainScript`2.Execute()
00000031 6.70576859 [2344] Error in AudioChainScript:
00000032 6.70576859 [2344]
00000033 6.70576859 [2344] Object reference not set to an instance of an object.
. . . and so on. It spams this.
00000311 7.86265707 [2344] Warning: ComputeContext(806126304) leaked!
00000312 7.86511326 [2344] Information: Dispose ComputeCommandQueue(634353200) in Thread(2).

[Zachs]

Thank you for the newest version!

This is what I now get in DebugView on a fresh default config with any AudioScript enabled (I apologise if this is redundant for you):

Ah I forgot to fix that.

Can you try it with the latest source version from GitHub?
I'll probably make another release once you've tested it out.

There's instructions on MPDN Extensions GitHub page on how to use the source version if you are not familiar with it.

** I've released extensions v1.26.1 which should fix the problem.

Let me know how you go.

[cascardian]

Works beautifully now, many thanks!

[Zachs]

Fantastic!

[ryrynz]

Tried updating via the player.

http://i.imgur.com/E3eoGJo.jpg

[Zachs]

@Belphemur any idea what's wrong?

[mecedo]

Can someone explain how to use Belphemur's subtitle drag-and-drop extension? I can't handle subtitles by mouse. I need the possibility to shift subtitles in vertical. The best option will be keyboard shortcuts (maybe it should be implemented in XySubFilter?)

[Shiandow]

Can someone explain how to use Belphemur's subtitle drag-and-drop extension? I can't handle subtitles by mouse. I need the possibility to shift subtitles in vertical. The best option will be keyboard shortcuts (maybe it should be implemented in XySubFilter?)

Are you trying to reposition subtitles by dragging them?

I'm pretty sure the subtitle drag / drop extension simply loads subtitle files when you drop them on MPDN's screen (similar to what you can do with video files and URLs)

I don't think the subtitles can be shifted at the moment, and I'm reasonably sure it can't be added as an extension.

You could try adding it as a feature request in MPDN's bug tracker (https://mpdn.zachsaw.com/bug/view_all_bug_page.php).

[mecedo]

Are you trying to reposition subtitles by dragging them?

Yes, you're right. My misunderstanding :)

[Anima123]

Shiandow, have you improved the soft algorithm within SuperRes in the latest version? I just found it was pleasing to set 'softness' to 0.05, together with the default SuperRes settings.

[Shiandow]

I don't recall any huge changes, but maybe it combines well with some of the other changes (perhaps the Sinc-Jinc prescaler)? Not sure if you're going to see much of a difference with softness set to 0.05 though.

[Belphemur]

Tried updating via the player.

http://i.imgur.com/E3eoGJo.jpg

Any content in particular that easily shows the flickering with Fluid Motion in previous releases?

We switched the MPDN release server to HTTPS, and the downloader included in the installer can't deal with it. I updated the installer as soon as I encountered the problem, but we need to release a new version of the extension to fix the problem.

For now, you need to download both installers manually:
MPDN X64 (https://mpdn.zachsaw.com/Latest/Installers/MediaPlayerDotNet_x64_Installer.exe)
MPDN X86 (https://mpdn.zachsaw.com/Latest/Installers/MediaPlayerDotNet_x86_Installer.exe)

Extensions (https://github.com/zachsaw/MPDN_Extensions/releases/download/1.26.1/MPDN-Extensions_v1.26.1_Installer.exe)

You need to run both of the installer, in any order.

[Anima123]

Use NNEDI3 (both version) together with SSIM result in a blackscreen, ctrl-K says there are error in render chain script.

Super-xBR is fine though. Once I thought it was the chroma algorithm within NNEDI3, but switch to another one doesn't fix the error.

Edit: BTW, I was using the latest render script commit 833bb8d.

Just checked with the latest release 1.26.1, the same error.

Edit2: It seems NNEDI3 is not cooperating with SSIM, other downscaling algorithms (didn't test them all) seem OK.

[aufkrawall]

Wow, the new super-xbr version is a huge improvement. It has far less problems with ringing than the old one, looks much cleaner.

I can confirm blackscreen issues with NNEDI3.

[Shiandow]

Figured out what causes the black screen. And added a method for detecting those kind of problems.

In future you'll see an "Uncompiled Filter." error, which probably doesn't tell you much, but I'll know where to look. This error will also trigger in cases where the problem isn't as visible, which might be a bit inconvenient but will ensure that these kind of problems are captured earlier.

[mlauzon]

I get a weird circular banding on HD live action movies & TV shows where it is colours of the scene, but it doesn't seem to happen on HD anime/cartoons or just SD cartoons.

How can I fix this, if possible? Because it doesn't happen while using other media players.

[Zachs]

I get a weird circular banding on HD live action movies & TV shows where it is colours of the scene, but it doesn't seem to happen on HD anime/cartoons or just SD cartoons.

How can I fix this, if possible? Because it doesn't happen while using other media players.
Are you using any render scripts? What sort of circular banding? Some samples would help.

[mlauzon]

Are you using any render scripts? What sort of circular banding? Some samples would help.

I don't know how to use any of those, everything is default.

I can't provide any examples, I've tried taking a print screen screencap, and all I get after pasting it to MS Paint is a black screen.

[Zachs]

How about a sample so we could try and reproduce the problem?

There's not much to go on with what you described.

[Zachs]

Figured out what causes the black screen. And added a method for detecting those kind of problems.

In future you'll see an "Uncompiled Filter." error, which probably doesn't tell you much, but I'll know where to look. This error will also trigger in cases where the problem isn't as visible, which might be a bit inconvenient but will ensure that these kind of problems are captured earlier.

Released as v1.26.2. Cheers.

[mecedo]

I've observed the following issue: If computer starts in 23.976Hz mode and we switch to another mode, i.e. 30Hz then every 30fps movie played in windowed mode plays with horrible stuttering. MPDN in Ctrl+J stats shows Composition Rate 23.976Hz. After switching to fullscreen movie plays fluently. But not in windowed. The only way to fix it is to restart operating system. But if we change refreshrate to another mode then problem occurs again and only restart helps.
Did anyone observed a similar issue?

[aufkrawall]

When there is downscaling after SuperRes needed, it uses the scaler selected in MPDN main application:
http://abload.de/thumb/srdownscaleynsjp.png (http://abload.de/image.php?img=srdownscaleynsjp.png)
Is there a way to use a render script shader for this? Maybe add a SSIM checkbox in SuperRes settings?

Btw: I'm still not fully happy with the deband filter. Could someone please port MPV's deband filter to HLSL? It's already ported to ReShade, if this is any help:
https://github.com/crosire/reshade-shaders/blob/master/ReShade/Shaders/JPulowski/Deband.fx

[Shiandow]

When there is downscaling after SuperRes needed, it uses the scaler selected in MPDN main application:
http://abload.de/thumb/srdownscaleynsjp.png (http://abload.de/image.php?img=srdownscaleynsjp.png)
Is there a way to use a render script shader for this? Maybe add a SSIM checkbox in SuperRes settings?


You can change the prescaler to "SuperXbr + SSIM" to achieve that.


Btw: I'm still not fully happy with the deband filter. Could someone please port MPV's deband filter to HLSL? It's already ported to ReShade, if this is any help:
https://github.com/crosire/reshade-shaders/blob/master/ReShade/Shaders/JPulowski/Deband.fx

In what way is MPV's debanding filter better than MPDN's?

[aufkrawall]

You can change the prescaler to "SuperXbr + SSIM" to achieve that.

I can't find it:
http://abload.de/thumb/srnds1m.png (http://abload.de/image.php?img=srnds1m.png)
It's not in the list of all available scalers either.


In what way is MPV's debanding filter better than MPDN's?
It's working very well for me in games, while MPDN's deband filter often misses quite some banding in videos compared to madVR. So I thought it might not hurt to have the option to try it.

Btw: Now that super-xbr has far less problems with ringing, it might make sense to add it also as a chroma scaler?

[lanzorg]

Online streaming, pipe input, RAR archives, DVD playback


Does anyone know how to launch MPDN from Firefox or Chrome with streaming websites like YouTube?

[Shiandow]

I can't find it:
http://abload.de/thumb/srnds1m.png (http://abload.de/image.php?img=srnds1m.png)
It's not in the list of all available scalers either.

No, you'd have to add it; if you press "modify list" you can add whatever prescaler you want, which can be as complicated as you like. Since there are increasingly more options it's not really doable to add every single possible method, but I'm always open to better options to put in the default list.


It's working very well for me in games, while MPDN's deband filter often misses quite some banding in videos compared to madVR. So I thought it might not hurt to have the option to try it.

Well, the method is quite simple so I can give it a try. I might try to improve a bit on the method. FWIW I currently see the biggest banding problems when the source is pretty much 'broken', MadVR might be a bit more aggressive in those cases, but in exchange MPDN's method can be used at pretty much full strength without completely obliterating the image.


Btw: Now that super-xbr has far less problems with ringing, it might make sense to add it also as a chroma scaler?

I'd have to see if that's possible, but for now I'd rather focus on the new reconstructive Bilateral method, which seems rather promising (still need some kind of 'limiter' to prevent artefacts, but other than that the quality is impressive).

[aufkrawall]

No, you'd have to add it; if you press "modify list" you can add whatever prescaler you want, which can be as complicated as you like. Since there are increasingly more options it's not really doable to add every single possible method, but I'm always open to better options to put in the default list.

I can add SSIM downscaler there, but then SuperRes isn't used at all according to OSD and it's still downscaled with MPDN internal downscaler.


Well, the method is quite simple so I can give it a try. I might try to improve a bit on the method. FWIW I currently see the biggest banding problems when the source is pretty much 'broken', MadVR might be a bit more aggressive in those cases, but in exchange MPDN's method can be used at pretty much full strength without completely obliterating the image.

I think this would definitely make sense, since even "HQ" sources often suffer severe banding.
When you set the threshold of your deband filter to a very high value, this kind of banding isn't reduced as much as with madVR deband filter, but the image gets a softer look in general too.



I'd have to see if that's possible, but for now I'd rather focus on the new reconstructive Bilateral method, which seems rather promising (still need some kind of 'limiter' to prevent artefacts, but other than that the quality is impressive).
Looking forward to it. :)
With cartoons, there are currently often errors which are still easy to spot.

[foozoor]

I use my laptop/ultrabook very often and sometimes I watch some videos.
Could you share a preset for the Intel HD 530 integrated GPU?

[Shiandow]

I can add SSIM downscaler there, but then SuperRes isn't used at all according to OSD and it's still downscaled with MPDN internal downscaler.

The OSD system might be having some trouble (making it more failproof has been on my todo list for a while) or something is wrong. As far as I can tell the this script (http://pastebin.com/1BJZJ4jK)should do what you want.


Looking forward to it. :)
With cartoons, there are currently often errors which are still easy to spot.

That goes both ways by the way, if you know where to look 'non-reconstructive' methods cause errors that are quite easy to spot as well ;). One of the easiest to spot is chroma bleeding (http://screenshotcomparison.com/comparison/164757), but sometimes it's simply loss of detail (http://screenshotcomparison.com/comparison/164756).

[aufkrawall]

The OSD system might be having some trouble (making it more failproof has been on my todo list for a while) or something is wrong. As far as I can tell the this script (http://pastebin.com/1BJZJ4jK)should do what you want.

Oops, mea culpa. I totally didn't consider that a script chain can be added as a prescaler option.
Works well and OSD reports it correctly.


That goes both ways by the way, if you know where to look 'non-reconstructive' methods cause errors that are quite easy to spot as well ;). One of the easiest to spot is chroma bleeding (http://screenshotcomparison.com/comparison/164757), but sometimes it's simply loss of detail (http://screenshotcomparison.com/comparison/164756).
Agreed. However, imho such kind of errors is less obtrusive than e.g. bright colored dots between areas of different color, like in cartoons.

[Shiandow]

Agreed. However, imho such kind of errors is less obtrusive than e.g. bright colored dots between areas of different color, like in cartoons.

Those kinds of artefacts should have stopped mostly with the new version (1.26.2), if not then something is wrong. The remaining artefacts are mostly colour not following edges correctly (usually when there's a dark edge next to a red area).

[aufkrawall]

The remaining artefacts are mostly colour not following edges correctly (usually when there's a dark edge next to a red area).
Yep, that describes it well.

[huhn]

Well, the method is quite simple so I can give it a try. I might try to improve a bit on the method. FWIW I currently see the biggest banding problems when the source is pretty much 'broken', MadVR might be a bit more aggressive in those cases, but in exchange MPDN's method can be used at pretty much full strength without completely obliterating the image.

so the debanding algo is changed now?

in the past it was powerful but it could easily remove a huge deal of details unlike madVR f3kd based deband filter.

[Zachs]

so the debanding algo is changed now?

in the past it was powerful but it could easily remove a huge deal of details unlike madVR f3kd based deband filter.
You might've remembered it the wrong way around.

[huhn]

You might've remembered it the wrong way around.

i did the test my self and posted it here so now. it was at the time the deband filter was still in part of madVR that some time ago! i can search for the post or at least the disc if you like.

[Zachs]

i did the test my self and posted it here so now. it was at the time the deband filter was still in part of madVR that some time ago! i can search for the post or at least the disc if you like.
I can't remember the earliest implementation of the filter but it's been like this for a long time i.e. preserves more details but not as strong.

[huhn]

found the post.

http://forum.doom9.org/showpost.php?p=1725141&postcount=30693

it extremely terrible in motion.

i can cut a sample later but this will take time i don't even know which season this was.

[Shiandow]

found the post.

http://forum.doom9.org/showpost.php?p=1725141&postcount=30693

it extremely terrible in motion.

i can cut a sample later but this will take time i don't even know which season this was.

That does seem to be an earlier implementation of the algorithm.

Might want to give it another try with the new version on default settings. Although that image looks like it could be 'broken' in the sense that you have to choose between representing the source accurately, or removing the banding.

[huhn]

got this crash after first time starting with a complete reset of all setting.

TITLE: SharpDX Error
------------------------------

An unexpected error 'SharpDX.SharpDXException' has occurred.

------------------------------
ADDITIONAL INFORMATION:

HRESULT: [0x887A0005], Module: [SharpDX.DXGI], ApiCode: [DXGI_ERROR_DEVICE_REMOVED/DeviceRemoved], Message: The GPU device instance has been suspended. Use GetDeviceRemovedReason to determine the appropriate action.
(SharpDX)

------------------------------
BUTTONS:

&Abort
------------------------------


it was most likely running out of vram still this is no reason to crash i guess.

[huhn]

That does seem to be an earlier implementation of the algorithm.

Might want to give it another try with the new version on default settings. Although that image looks like it could be 'broken' in the sense that you have to choose between representing the source accurately, or removing the banding.

yes it is clearly changed.
and it doesn't harm the truck anymore.

this is with 1.0
http://abload.de/img/truckxoad8.png

[aufkrawall]

MPDN's deband with a very high strength looks a bit like a deblocking filter. I notice this when e.g. watching a news video, the face of the newsspeaker becomes a bit too blurry.

[Shiandow]

FYI I think the 'power' setting doesn't really do anything at the moment. Should probably fix that at some point.