LAV Filters - DirectShow Media Splitter and Decoders

[TheShadowRunner]

How does it play when you remove the attached fonts?
Yes it works correctly (the signs are displayed), but I'd rather keep the original/intended font which works just fine with Haali ^^;
Edit: if you test the sample on your side do you have the issue?

[LigH]

MPC-HC 1.7.10-207 x64 with internal LAV 0.68.0.28: No issues.
MPC-HC 1.7.10-269 x64 with internal LAV 0.68.1.33: No issues.

Subtitles appear in the attached font face "Neo Sans" like in your screenshot. Overriding the ASS font with "Standard style" as defined in MPC-HC's options (default: Arial) works well too.

Attached are OpenType fonts. Not sure if their support in Windows XP was limited in some surprising details...

I did not test with external filters and system-wide installed LAV Filters. But I would expect that if a splitter supports extracting and temporarily registering an attached font at all, then it should be supported as a whole. Hard to imagine reasons to exempt single characters.

[uneedme]

I dont know how to describe it...... so i say tearing......

did some test on it after the post...

potplayer:


lavfilter decoder:

vmr9 -rare occasions
EVR(CP) -always
d3d9 -always
madshi randerer -never
opengl randerer -never


ffmpeg decoder:

vmr9 -rare occasions and less effect
EVR(CP) -rare occasions and less effect
d3d9 -rare occasions and less effect
madshi randerer -never
opengl randerer -never


After did a "Platform update", i dont know what that is......, the problem is gone. Looks like it is Windows' some hardware usage (graphics stack) glitch......been fixed by patches......

After I mounted the new 2nd-hand display card, I re-installed the win7...So the system might roll back to the glitchy conditions...And this is the first time I heard about a big patch called "Platform update"......

:mad:

https://www.microsoft.com/en-us/download/details.aspx?id=36805

[TheShadowRunner]

Thanks for your feedback.
Attached are OpenType fonts. Not sure if their support in Windows XP was limited in some surprising details...
It cannot be.. otherwise the flaw would be there for Haali Splitter as well(?)

But I would expect that if a splitter supports extracting and temporarily registering an attached font at all, then it should be supported as a whole. Hard to imagine reasons to exempt single characters.
Yes it's an odd bug indeed..

[nevcairiel]

You don't happen to have a font with the same/similar name installed in the system already, do you?
For some reason it won't let you install a new font properly if there is already one with the same basename, for example. It can result in rather odd looking issues. Not sure if the font registering or the font usage in MPC-HC is to blame.

As mentioned above, the font name is "NeoSans". Anything with that name in your Windows font list?

There isn't exactly much one can do wrong when registering fonts. Take the font blob from the mkv and give it to some windows API.

[TheShadowRunner]

You don't happen to have a font with the same/similar name installed in the system already, do you?
No, no similar font is installed on the system.

As mentioned above, the font name is "NeoSans". Anything with that name in your Windows font list?

No.. :X

There isn't exactly much one can do wrong when registering fonts. Take the font blob from the mkv and give it to some windows API.
Hmm I'm not sure I follow, NeoSans isn't installed, it's only available as attachment in the MKV. If I remux without it, I fall back to the default (and ugly) font.
But everything is ok when using Haali hence my surprise.
It seems it's LAV Splitter itself that isn't passing everything it should to Directvobsub (latest 1.7.9.0).

[Telion]

I have an XP, too, so I tried this with LAV, MPC-BE, XySubFilter, madVR, no NeoSans installed and got the same behaviour.
But since MPC-BE can load external fonts (from the "fonts" subfolder of a folder with video) I tried to extract the fonts there and remux the original mkv without them - the problem is gone. Then I tried to play the original mkv while the "fonts" subfolder is still there - there is no problem, either. So I conclude there is something that LAV Splitter is doing wrong since MPC-BE font loading does it right.

[nevcairiel]

Loading Fonts from memory and from files is probably the big difference. Maybe Haali writes it out into a temporary file to avoid that issue? Sounds a bit ugly to manage some temp files for that.
I could probably hack something up for testing, a basic version shouldn't be too hard, just need to think about the issues of avoiding leaving behind orphan font files in the temp folder everywhere.

[Telion]

Loading Fonts from memory and from files is probably the big difference. Maybe Haali writes it out into a temporary file to avoid that issue?
Seems you are right. IIRC Haali does indeed extract fonts to %TEMP% folder. TheShadowRunner, can you please confirm this? And I checked the MPC-BE's internal mkv splitter - it doesn't write temp files and does have this issue.

[Sarasa]

Tested the sample "samplecbbp.mkv" with MPC-HC 1.7.10-269 (86/x64), LAV 0.68.1.35 (Install version), madVR 0.90.24, XySubFilter 3.1.0.746
I don't have the bug and I don't have any "NEO sans" font installed

Xp Limitation ?

[clsid]

Yes, works fine on Win7. The temp file workaround should only be needed on XP. Has anyone tried on Vista?

[TheShadowRunner]

Seems you are right. IIRC Haali does indeed extract fonts to %TEMP% folder. TheShadowRunner, can you please confirm this? And I checked the MPC-BE's internal mkv splitter - it doesn't write temp files and does have this issue.
Yes you guys have found the culprit!
(with Haali) I have indeed some temp files created in:
C:\Documents and Settings\user\Local Settings\Temp
with name such as mkf4B7.tmp.
I opened one in a hexeditor and bingo I can see it's related to the font (OTTO, Neo Sans is a trademark of AGFA ,etc..). It's also deleted once I exit the player (ZP).
The mystery is solved, thanks for testing.

[VictorLS]

My prayers were answered: my former (until 1991 in USSR) and current (from march 2014 in Russia) compatriot, true master of programming and developer of MPC-BE Aleksoid1978 has implemented h265 10bit (and VP9 8bit I not tried yet) hardware decoding (using philipl's investigation https://devtalk.nvidia.com/default/topic/965840) in LAV Video Decoder 0.68.1.35git via CUVID in one or two days and it works great in WinXP. I can watch 12 UHD channels in WinXP with 50Hz settings absolutely smooth in DVBViewer. I can receive also one 19e2 channel VR Demo h265 Main@L6@Main 10000x1700 25fps 4:2:0 8bit without hardware acceleration, of course.
I want repeat to nevcairiel from here http://forum.doom9.org/showthread.php?p=1784923#post1784923: sometimes much easier to do something then seems simply say no. But I want say to nevcairiel THANKS for LAV Filters once more.
Link for patch https://yadi.sk/d/5NtWmfjPyoQZe (64kB)
Link for installer of patched LAV Filters https://yadi.sk/d/M5KOnw9YyoQWo (10MB)
PS. I will try patched LAV Filters in Win10x64 at weekend too - may be it will work in it.

[clsid]

Tried that patch, but HEVC10 decoding is broken and crashes my driver (373.06).

VP9 Profile 0 is working.
VP9 Profile 2 does not fall back to software decoding. sample (https://mega.nz/#!hJdFEIBI!uSOjZtkkjIVYSfqD9aSfONf1yq__uQvlsf47pCtFvdQ)

[davidsama]

VictorLS, i want repeat to nevcairiel from here http://forum.doom9.org/showthread.ph...3#post1784923: sometimes much easier to do something then seems simply say no.
I read that line as sometimes much easier to do to what i tell you to do and do not argue with me about it then seems simply say no.

[davidsama]

Clsid, Have you tried https://www.nvidia.com/download/driverResults.aspx/111032/

[VictorLS]

I read that line as sometimes much easier to do to what i tell you to do and do not argue with me about it then seems simply say no.
You are some right - some because nevcairiel's argue about philosophy not enough for me - nothing real argue that possibility of h265 10bit decoding in LAV Video Decoder through CUVID will worse for somebody (even if so there is always choice to set None hardware decoding as happens automatically now). I said many times - choice must be given for people and they choose the best. For example, question about artifacts: nevcairiel give not choice in LAV Video Decorer as I mentioned in some pages in this thread before. I know this isn't difficult to realize, it will not take much space in LAV Video Decoder's interface but his philosophy prohibits do that again.
PS. I don't see difference between cuvid and DXVA2 on my 8bit monitor. People with 9,10bit monitors have choice to choose DXVA2. I think all satisfied...
This is my point of view.

[VictorLS]

VP9 Profile 0 is working.
Yes, but 4K 60fps plays with some judder instead of smooth 1080p 60fps and 4K 30fps - may be my GTX750 is too weak for 4K 60fps VP9 profile 0.
Tried on The Curvature of Earth 4K 60FPS-UceRgEyfSsc.webm downloaded with youtube-dl https://www.youtube.com/watch?v=UceRgEyfSsc --f 315 (313, 303)
PS. Just knew about youtube-dl. It's great console utility - I have no much internet at home.
VP9 Profile 2 does not fall back to software decoding.
Yes, with both The World in HDR.mkv and The Redwoods.mkv
Has anybody VP9 Profile 1 sample?

[VictorLS]

may be my GTX750 is too weak for 4K 60fps VP9 profile 0
I was wrong in this case - MadVR has big advantage over VMR7 and OverlayMixer in WinXP and plays that 4K 60fps video smooth instead of old renderers (first time see that).
GTX750v2 is great videocard.
PS. Samsung_HDR_Wonderland.ts http://forum.doom9.org/showthread.php?p=1785832#post1785832 (monitor's frequency set to 47,952Hz in nVIDIA Control Panel) in WinXP with patched by Aleksoid1978 LAV Video Decoder 0.68.1.35-git plays without dropped frames with MadVR. In Win7x64 frames are always dropped (both with CUVID and with DXVA2) with MadVR but with EVR plays smooth as in WinXP with MadVR and OverlayMixer.

[clsid]

Clsid, Have you tried https://www.nvidia.com/download/driverResults.aspx/111032/I haven't tried recent drivers, since those are full of bugs.

If the code requires a minimum driver version, then it should check for that. Crashes are unacceptable.

[Sarasa]

Don't use the "375.86 WHQL" they have a bug
Low memory clocks on certain factory-overclocked Pascal cards.

use the new one "375.95 Hotfix" ;)
Can be found here > http://nvidia.custhelp.com/app/answers/detail/a_id/4260

[VictorLS]

VP9 Profile 2 does not fall back to software decoding.
Redownload and install patched LAV Filters https://yadi.sk/d/M5KOnw9YyoQWo (10MB)
Aleksoid1978 said VP9 Profile 2 (and other video that CUVID can not accelerate) must fall to software decoding now.

[kathykit]

Don't use the "375.86 WHQL" they have a bug
Low memory clocks on certain factory-overclocked Pascal cards.

use the new one "375.95 Hotfix" ;)
Can be found here >

:thanks: back drop
I am just try download that, but my computer just warn me.
maybe thats the reason.

[clsid]

Redownload and install patched LAV Filters https://yadi.sk/d/M5KOnw9YyoQWo (10MB)
Aleksoid1978 said VP9 Profile 2 (and other video that CUVID can not accelerate) must fall to software decoding now.
VP9 fallback works now. Do you have the patch too?
Hevc10 looks like this: pic (http://i65.tinypic.com/1zxmmc7.png)

[nussman]

I think Aleksoid1978 should explain this patch here.

[VictorLS]

VP9 fallback works now.
Yes, in WinXP too.
Do you have the patch too?
I've asked Aleksoid1978's permission to give link.
Hevc10 looks like this
Upload your screenshot on other site, please. For example as me here http://forum.doom9.org/showthread.php?p=1785832#post1785832

[Aleksoid1978]

Here is a patch - https://yadi.sk/d/laQCuqFAyso7D

[LigH]

Upload your screenshot on other site, please.

I mirrored it there (http://frupic.frubar.net/shots/34634.png).

[VictorLS]

I mirrored it
Thanks

[VictorLS]

I've formated C: and installed there Windows 10 Anniversary Update 1607 build 14393 Enterprise x64 from 14393.0.160715-1616.rs1_release_CLIENTENTERPRISE_VOL_x64fre_ru-ru.efr.iso from 4GB flash then install update windows10.0-kb3200970-x64.msu and 375.95-desktop-notebook-win10-64bit-international.hf.exe
I was twice well surprised:
1. Win10's Movies&TV plays many (LAV Splitter opens much more my SAT recordings) of h265 8 and 10bit and VP9 with DXVA2 (because I see artifacts in h264 RussiaHD.ts http://forum.doom9.org/showthread.php?p=1784923#post1784923) without installing any filter.
2. I not see artifacts with CUVID in Aleksoid1978's patched filters (I did regsvr32 them - not install from his exe) in RussiaHD.ts and other such files as in Win7 happens.
I haven't seen any unexpected playback issues (including clsid's http://frupic.frubar.net/shots/34634.png) while playing more than 200 h265 4K files with CUVID (many people here told CUVID not work in Win10 at all - it isn't true in my experiment).
But Win10 is still bad for me:
1. It hangs up in a couple of second after booting begins. I have to press Reset button of my PC - second boot successful but cause HDD check in a future;
2. Not work my old radioremote control ATI Remote Wonder (may be someone knows how to?);
3. Not work my favorite version of DVBViewer;
4. Not friendly interface (for example do Ctrl+Shift instead of Alt+Sift to switch languages is some difficult) and many many other causes...
PS. With CUVID hardware decoding in Win7 I can use again my favorite KMPlayer 4.0.7.1 by 7sh3 without advertising and not send what file is played to someone. IMHO it best with ATI Remote Wonder and it has almost ideal interface (but in KMP I can not set native size of 4K video to see 1/4 of square on my FullHD monitor as MPC-HC can), because DXVA2 (copy-back) glitches sometimes and LAV's DXVA2 (native) not work in KMP at all.

[gpower2]

1. It hangs up in a couple of second after booting begins. I have to press Reset button of my PC - second boot successful but cause HDD check in a future;
Usually a sign that one of the disks is failing, or some other external device is blocking the boot sequence (I've seen that to a PC that had 2 keyboards connected at the same time, one in PS2 and one in USB).
You should check with as little peripherals as you can in order to pinpoint the exact problem in your system.
2. Not work my old radioremote control ATI Remote Wonder (may be someone knows how to?);
You should check the Hardware IDs at the Details tab from Device Manager for the remote control device and check Google in case there is a driver for Windows 10 for that specific device.
3. Not work my favorite version of DVBViewer;
You should check ProgDVB, it has frequent updates and it is written by a Russian developer, so you could have a lot of feedback for your local channels ;)
4. Not friendly interface (for example do Ctrl+Shift instead of Alt+Sift to switch languages is some difficult) and many many other causes...
Uhm, you can change that key combination from the control panel???
PS. With CUVID hardware decoding in Win7 I can use again my favorite KMPlayer 4.0.7.1 by 7sh3 without advertising and not send what file is played to someone - imho it best with ATI Remote Wonder and it has almost ideal interface (but in KMP I can not set native size of 4K video to see 1/4 on my FullHD monitor as MPC-HC can), because DXVA2 (copy-back) glitches sometimes and LAV's DXVA2 (native) not work in KMP at all.
I find latest nightly build of MPC-HC to be exceptionally good with every video I feed to it (having disabled all the internal splitters and decoders in order to use latest ofiicial LAV), even better that MPC-BE or other commonly used players out there.

[nevcairiel]

2. I not see artifacts with CUVID in Aleksoid1978's patched filters (I did regsvr32 them - not install from his exe) in RussiaHD.ts and other such files as in Win7 happens.

Thats because the HQ DXVA mode doesn't work on Windows 10, and it falls back to the CUDA mode - which is just one problem CUVID has on Windows 10, the other one is that it can deadlock when OpenCL is used, ie. by madVR, which is really quite an annoying bug.

[VictorLS]

check with as little peripherals as you can in order to pinpoint the exact problem in your system.
I'll check.
check Google in case there is a driver for Windows 10 for that specific device.
Even if I will find working driver, there must be program for ATI Remote Wonder. Even Win8x64 last official version worked not stable.
check ProgDVB
I respect Prog (for example he implement Maincocept HEVC Video Decoder support instead of nevcairiel, but ProgDVB on .NET and works very slowly compared with DVBViewer. Although ProgDVB has some advantages: for example, it can receive multistream transponders.
for your local channels
Artifacts is a global problem of DXVA on nVIDIA and AMD videocards in Win7 as minimum, but nVIDIA has CUVID to decide problem instead of AMD. nVIDIA's Level2 support hasn't answered me about artifacts yet )
you can change that key combination from the control panel???
I did it in one minute but it is so deep compared with Win7 ad XP.
I find latest nightly build of MPC-HC to be exceptionally good with every video I feed to it, even better that MPC-BE or other commonly used players out there.
I believe but what about remote control and user friendly interface?
having disabled all the internal splitters and decoders in order to use latest ofiicial LAV
Why so difficult - I just replace all 12 files in MPC-HC\LAVFilters(64) folder ;)
But I use Aleksoid1978's patched filters.

Thats because the HQ DXVA mode doesn't work on Windows 10, and it falls back to the CUDA mode - which is just one problem CUVID has on Windows 10
Thanks for explanation.
the other one is that it can deadlock when OpenCL is used, ie. by madVR, which is really quite an annoying bug.
I haven't tried MadVR in Win10 yet. I hope will be time for that before I uninstall Win10 )

[TheShadowRunner]

nev, should I report the font / temp file (missing characters bug) on the GitHub project page?

[gpower2]

Even if I will find working driver, there must be program for ATI Remote Wonder. Even Win8x64 last official version worked not stable.
You could use a generic remote control application that lets you assign specific application macros. ;)
I respect Prog (for example he implement Maincocept HEVC Video Decoder support instead of nevcairiel, but ProgDVB on .NET and works very slowly compared with DVBViewer. Although ProgDVB has some advantages: for example, it can receive multistream transponders.
Well ProgDVB uses ngen during installation, so .NET code gets compiled in native code, so it's not really a matter of .NET vs native code.
I believe but what about remote control and user friendly interface?
User friendliness is quite subjective. As for the remote control capabilities, I strongly suggest to use a generic application.
Why so difficult - I just replace all 12 files in MPC-HC\LAVFilters(64) folder ;)
It's not wise to replace those files, since a simple API change in the filters can crash the whole player. ;)

[VictorLS]

You could use a generic remote control application that lets you assign specific application macros. ;)
I can but better simply stay on quite stable WinXP )
I don't want waste much time for this.
Well ProgDVB uses ngen during installation, so .NET code gets compiled in native code, so it's not really a matter of .NET vs native code.
Thanks, I didn't know. By the way for mulistream transponders I use old good fast (but not stable) AltDVB with LAV Video (CUVID acceleration) and VMR7 both in WinXP and in Win7.
User friendliness is quite subjective.
I agree. User friendliness of KMP the best on my own. But I have to mention KMP works with filters and renderers worse then MPC-HC(BE) or POT Player.
As for the remote control capabilities, I strongly suggest to use a generic application.
If you want help me to set up generic application to remote control MPC-HC please write PM - it's offtop here.
It's not wise to replace those files, since a simple API change in the filters can crash the whole player. ;)
Theoretically you are right, but practically I've never seen any problem both in KMP and in MPC-HC.

PS. I tried disconnect all connected to my PC devices excluding SAT-card but including keyboard and mouse and second HDD - Win10 still hangs up (

[Nintendo Maniac 64]

Do your laptop not support 1080p hardware acceleration? Can you provide hardware (model of laptop, CPU, GPU if present) and software (OS and player with renderer you use or tried) configuration? Although in any case you have to manually switch from None to DXVA2 in settings of LAV Video Decoder depending on resolution of played video and on battery/on electrical line.
PS. You are lucky - you have this possibility - many people of the world who see artifacts with DXVA can not switch it of in LAV Video Decoder and they don't know about LAV CUVID Decoder. They have to switch off hardware acceleration because they have no choice (but sometimes with weak CPU it is impossible to decode smooth 1080p without hardware acceleration). I hope for them that nevcairiel will add possibility to switch off DXVA Processing in LAV Video Decoder.

I did some quick-and-dirty testing and I think "None" actually doesn't hit my CPU anywhere near as hard as I thought it would; I think I'm too used to web browsers which not only usually have less efficient decoders but also can have fancy-pants javascript-heavy video players (hello twitch.tv!).

Not only that, but because it's an older laptop that uses discrete graphics rather than integrated, it seems that doing GPU decoding (DXVA or CUVID) pulls more power than one would expect as well.


I'll try using "None" for the next few weeks or so and see how it goes.

[CruNcher]

Depends on the Decoder sure the Discrete Decoder can drain the battery very fast Quicksync i guess you refer to will be always more energy efficient, then going the discrete path of AMD or Nvidia.

And yes depending on the Decoder the CPU can be in certain Scenarios more efficient as well, for example in a Hybrid Decoding Scenario you mostly allways better of using the CPU decoder for the Web :)

But of course if the other side doesn't care about efficiency in their Render Stack that will go to hell pretty fast ;)

[har3inger]

I have a quick question about matrix encoding in LAV audio for 5.1-> 2.0. If I select either dolby pro logic I or II, the audio center moves to the right when played back regularly on headphones. Is this intended behavior? More specifically, the center channel stays center after downmixing, but the music from the other channels all moves to the right.

[nevcairiel]

Matrix encoding is only useful if you have a decoder on the other end that understands these matrix encodings. For playing on a stereo system or headphones, use regular downmixing.

[LigH]

Well, in my experience, Pro-Logic downmixes should be quite compatible to common stereo playback systems. You may not have a really distinct front/back separation, but the phase shifted weighted mix of surround channels into left and right channels still gives an impression of a less clearly located source.

A heavy weight on only one channel is certainly not the expected result, especially not when the source is mostly centered with only little surround impact; the center channel should be mixed equally to left and right if your downmix matrix is correct. Only surround channels should have a chance to get mixed with unequal weights.

[har3inger]

That's what I thought expected behavior to be. I'd assume that there is a bug with the way the matrix is being encoded currently? I have tried several samples where the audio clearly shifts to the right (even visible in the LAV sound channel volume tab) when switching from "none" to "pro logic II". Can anyone else reproduce to ensure it's not a config problem on my end?

I'm using matrix encoding for when I plug my computer into my TV with its 5.1 speaker setup, but leave it on for computer-only because of laziness.

[VictorLS]

I did some quick-and-dirty testing and I think "None" actually doesn't hit my CPU anywhere near as hard as I thought it would
You want to say: despite of "None" in LAV Video decoder some hardware acceleration are active? I think this is impossible.

[ryrynz]

You want to say: despite of "None" in LAV Video decoder some hardware acceleration are active? I think this is impossible.

He means to say that it's well optimized, as Nev has said earlier in the thread software decoding is preferred for most systems.. less issues and performance is very good.

[LigH]

To say it very generally:

Hardware chipset support speeds up the decoding process only if the decoder chip is faster than the CPU. Which is not anymore the default case. The current fast multi-core CPUs with SIMD instruction set extensions are equal or even superior to GPU decoders in many cases. Using hardware support is not generally an "acceleration".

[NikosD]

This is totally wrong.

I mean it is dead wrong.

Even from 2011, Intel's Sandybridge iGPU using QuickSync - a dedicated ASIC inside the GPU - was a lot faster than Core i7 CPU in all three codecs supported - H.264, MPEG2, VC-1/WMV3.

And the hardware decoder was present even in Core i3.

Nowadays, Intel's Skylake/Kabylake iGPU and Nvidia's Maxwell GM206/ Pascal have only increased the performance difference from the fastest Core i7 CPUs.

And the hardware decoders of Intel are present and activated even in Celerons/Pentiums since Haswell.

The latest variations of hardware decoders have added HEVC (8bit/10bit) and VP9 (mostly 8bit and Kabylake 8bit/10bit) with resolutions supported up to 8K.

Hardware dedoders are no match even for the fastest CPUs nowadays.

[VictorLS]

ryrynz, LigH, NikosD
I investigate hardware acceleration by videocards inserted in PCs with very weak CPUs since MPEG2 appeared and I agree with NikosD. On old lap top I'm sure it's must be more visible. It's a pity Nintendo Maniac 64 not answered my questions: Can you provide hardware (model of laptop, CPU, GPU if present) and software (OS and player with renderer you use or tried) configuration?
So I can't tell him what happens.

[LigH]

Okay, my scope in the previous reply was probably wrong. I remembered early Nvidia PureVideo generations being limited to common display framerates, they were an excecption to the generalized opinion that GPU decoding is "always" faster. And just displaying is also a different purpose than converting, where the result of the GPU decoding has to be transferred back to the RAM. :o

[nevcairiel]

It makes no difference how fast a decoder is for realtime playback if all options can play all your videos fast enough, however.

The key point is that even CPUs these days are plenty fast for FullHD H264 playback without any serious CPU usage, without requiring some i7 high-end CPU, so software decoding is a valid and viable option for a lot of people. Only if you go 4k or exorbitantly high bitrates it may no longer work - but thats far from what the majority actually plays.

[Aleksoid1978]

Latest NV CUvid with 375.xx drivers support P016 output and HEVC 12 bit decoding(not tested yet).