LAV Filters - DirectShow Media Splitter and Decoders

[huhn]

you should really disable SD hardware acceleration in this case (in my eye in general but what ever).

an SD stream doesn't really need any significant CPU power. software decoding has really good error handling and errors are common in broadcast.

[fxv300]

If history with AMD tells us anything, its their drivers. But I do not plan to buy such a card.

Hi, a lot of the issues of the problems with the drivers have already been fixed. Blog here (https://rx480.blogspot.co.uk/)

Re DXVA HW acceleration issues, I can now play SD HD and UHD with HW acceleration turned on. This may not be an RX480 problem after all so don't jump to any conclusions as yet.

DXVA (https://youtu.be/hCCBxKPOI84)Video

[nevcairiel]

Re DXVA HW acceleration issues, I can now play SD HD and UHD with HW acceleration turned on. This may not be an RX480 problem after all so don't jump to any conclusions as yet

If newer drivers fixed it, and it works now, then my conclusion would appear to have been correct and it was a driver problem? :)

[CruNcher]

Hi, a lot of the issues of the problems with the drivers have already been fixed. Blog here (https://rx480.blogspot.co.uk/)

Re DXVA HW acceleration issues, I can now play SD HD and UHD with HW acceleration turned on. This may not be an RX480 problem after all so don't jump to any conclusions as yet.

DXVA (https://youtu.be/hCCBxKPOI84)Video

You should find another recording method the actual achieved framerate distracts a lot from the main problem you want to show especially in the case of DXVA.
Also the way you trying to show it is wrong you need to get to the running Lav Video Process in the Directshow Chain to show what is really happening.
That Dialog is not displaying the running Process Thread in the Chain but opening another Process that will always show as if no Decoder is active at all, because for that new running instance that's the truth it has no active connection to anything.

To achieve the lowest latency with Hardware acceleration try to record over the VCE on the DWM Surface in your Windiows 10 Case.

Broadcast Streams are more complex as huhn said and neither ATI was very good at resilience Decoding back then when they most of all targeted clean Blu-Ray Bitstreams.

UVD inside HD2400 RV610/RV630 UVD 1.0 was a total mess (Driver as well as the Firmware itself) in those regards no chance vs Nvidias VPX core stability in those times ;)

It was the time i got really frustrated with ATI and my change to Nvidia where they happily interacted with Consumers and Developers pretty open with for that time astonishingly low NDA barriers to even further improve the VPX core and it's spread everywhere and their initiation of VDPAU(CUVID), this also happened a lot over DOOM9 Feedback ATI was absolutely not interested a mayor management decision failure back then that surely bites AMD now in the ass improving everything backward.

[IanD]

Which is the last release to support XP SP2?

[manolito]

Nevcairiel does not "officially" support WinXP any more, but I can assure you that the current stable version 0.68.1 works nicely under WinXP SP2. (AFAIK this is not true for SP1)

Cheers
manolito

[Gleb Egorych]

Hi, nevcairiel,
I would like to return to the problem reported here: http://forum.doom9.org/showthread.php?p=1752214#post1752214
Shortly: LAV CUVID decodes the file with artifacts, while CoreAVC CUDA and LAV software (FFMpeg) do the job without artifacts. Interestingly that wanezhiling and clsid confirmed they had artifacts with LAV CUVID as well.
You wrote that sample played fine for you and it contained stream errors.

Here is a new sample, as the previous one it was recorded from SatTV: https://www.sendspace.com/file/hryhxw
Same problem for me, win7 x64, the newest geforce driver 372.70.

Maybe there is some workaround like you did for deinterlacing with the "Aggressive" option. Success of CoreAVC CUDA gives a chance.

[VictorLS]

Gleb Egorych, what version of CoreAVC with CUDA are you tried?
PS. I almost always use "Aggressive" deinterlacing in LAV Video Decoder. I hope to try your sample tomorrow.

[IanD]

Nevcairiel does not "officially" support WinXP any more, but I can assure you that the current stable version 0.68.1 works nicely under WinXP SP2. (AFAIK this is not true for SP1)

I tried installing 0.68.1 but it complains I need XP SP3 and refuses to install.

I have to go back to 0.59.1 (trial & error) before it will install on XP SP2.

[andyvt]

I received a sample file that causes a crash in avutil-lav-55.dll. The same file plays via ffplay.exe. LMK if you need more info.

Thanks

Sample (http://babgvant.com/downloads/samples/test-crash.zip)

[filler56789]

I tried installing 0.68.1 but it complains I need XP SP3 and refuses to install.

I have to go back to 0.59.1 (trial & error) before it will install on XP SP2.

Simply do not use the installer, just extract the 32-bit files from the installer,
and then run regsvr32 on the DirectShow filters (*.AX)

[clsid]

Simply do not use the installer, just extract the 32-bit files from the installer,
and then run regsvr32 on the DirectShow filters (*.AX)
The requirement is there for a reason. It uses Windows API functions only available in SP3. Manual registration will fail.

I received a sample file that causes a crash in avutil-lav-55.dll. The same file plays via ffplay.exe. LMK if you need more info.
Sample (http://babgvant.com/downloads/samples/test-crash.zip)
Works fine here with 0.68.1.31

[filler56789]

The requirement is there for a reason. It uses Windows API functions only available in SP3. Manual registration will fail.

So I stand corrected, sir. I never used LAV Filters with XP Service Pack 2, so I trusted what manolito said:

... but I can assure you that the current stable version 0.68.1 works nicely under WinXP SP2. (AFAIK this is not true for SP1)

[manolito]

Sorry I mixed up SP2 and SP3... :stupid:

The current LAVFilters do need XP SP3, but may I ask why you refuse to install SP3? I never heard of any issues which SP3 introduced.


Cheers
manolito

[rancorx2]

hi, is there any chance of getting an aspect ratio setting added to lav filters?

currently i use ffdshow for everything except hevc which i use lav for but it leaves me with black bars.

i would love to be able to stretch videso to fullsceen like you can do with ffdshow.

[clsid]

You can do that with MPC-HC.
menu > view > video frame > stretch to window

[VictorLS]

Gleb Egorych, your 2016-09-20 21-15-33.ts recorded from 1HD is played without artifacts in WinXP with LAV Video Decoder 0.68.1 CUVID (with last XP Cyberlink Video Decoder (PDVD Generic) and MPC Video Decoder and ArcSoft Video Decoder I see square artifacts along all file - DXVA is bad) but hangs up at the end of file in KMP 4.0.0.0 (in MPC-HC 1.7.10.236 all OK).
PS. If CUVID will decode h265 10bit - I have written to nVIDIA about it but have no answer ( - I don't need DXVA at all.

[akim]

nevcairiel
MPC-HC (LAV Filters) and mad-VR just a fairy tale thanks for that (can integrate them)
Friend implement or use the sane sound normalization in vlc (compressor) or potplayer even if they are incorrect
Another question is why there is no such option in the PAL SpeedDown player except Reclock, it is difficult or even something this amazing option
2-3 have to use one player to another normalization for PAL 25 to 23.976 compensation
One wants to use mpc-hc 64bit reclock but it does not support.
Thank you

[clsid]

Windows has an option called "loudness equalization". Google for it. Nevcairiel has already said several times he won't add such functionality to LAV.

The internal audio renderer of MPC-HC might one day get similar functionality as reclock. But not very soon.

[robl45]

You can do that with MPC-HC.
menu > view > video frame > stretch to window

I think the options is actually pan and scan scale to 16:9

[akim]

clsid
it is sad
I'll have to keep the 2-3 player (it is so inconvenient and wasteful)
loudness equalization - is not supported on my Audigy Rx (stereo speakers) Win 10
Thanks.

[Gleb Egorych]

VictorLS, CoreAVC 3.0.1, windows 7 x64 and geforce 372.70 for GTX660. What videocard do you have?

[VictorLS]

Gleb Egorych
What videocard do you have?
As I mentioned in this thread http://forum.doom9.org/showthread.php?p=1780356#post1780356 new GTX750 with h265 VLDecoding, http://forum.ixbt.com/topic.cgi?id=10:60902-42 and my Info there.
With LAV CUVID Decoder - DirectShow CUVID Video 0.13 your 1HD sample playing well with unchecked Use HQ DXVA Processing only. With checked it is horrible - worse then decoded by any DXVA1 decoder I tried including CoreAVC 3.0.1 with DXVA Accelaration, with CUDA and None plays well and even KMP not hangs up. This is one more thing why I (and others) want that CUVID (not DXVA) will decode h265 10bit https://devtalk.nvidia.com/default/topic/965840 in XP, 7x64, Linux .... But technical support of nVIDIA worse day by day.
So, it is DXVA issue with NTV+ (like your sample 1HD) channels - some people from http://viaccessfree.su/forum thought this issue due to many GOPs in stream http://webcache.googleusercontent.com/search?q=cache:BewI3O8OeawJ:http://viaccessfree.su/forum/showthread.php?p%3D765660%2Bdvbviewer+gop+site:viaccessfree.su&lr&hl=ru&ct=clnk
IMHO maximum that nevcairiel can do before CUVID will decode h265 10bit - implement mechanism for automatic switching between CUVID and DXVA2 in his LAV Video Decoder. But IMHO it is difficult begining from interface of decoder settings and nevcairiel never do what I ask (
PS. Due to my new videocard my asks about Mainconcept HEVC Decoder support in LAV Splitter and possibility of decoding only key frames in LAV Video Decoder are cancelled by me now )

[Gleb Egorych]

VictorLS, could you please test the sample in win7 as well?

Thanks to VictorLS, I tried retro LAV CUVID Decoder 0.13 and with disabled "Use HQ DXVA Processing" the sample is decoded perfectly. There is no such option in LAV Video, so I guess in LAV Video decoder it is disabled on XP and enabled on Vista+.

nevcairiel, is it possible to use HQ DXVA Processing for deinterlacing only? Accroding to your guide (https://1f0.de/lav-cuvid/guide/) the main purpose of the option is to access high quality deinterlacing engine.

[VictorLS]

Gleb Egorych
I use attachments settings of LAV Video Decoder and all work perfect both in WinXP and Win7x64 (except h265 10bit).
And I do not see sense of Win7x64 experiment - with CUVID WinXP and Win7x64 are the same (renderer doesn't matter in artifacts question). With DXVA2 I saw artifacts on many NTV+ channels (SD and HD) many times before in Win7x64 on EVR.

[Gleb Egorych]

Made a quick check with "slices" deinterlacing test:
LAV Video (CUDA), LAV CUVID Decoder with HQ DXVA checked and CoreAVC CUDA produce identical results, utilizing high quality deinterlacing.
Unchecking HQ DXVA in LAV CUVID Decoder clearly affects deinterlacing quality in a negative way.

So CoreAVC can get CUVID acceleration, HQ deinterlacing and does not produce artifacts on the sample (http://forum.doom9.org/showthread.php?p=1781141#post1781141). LAV requires HQ DXVA disabling for artifact-free playback of the file.

All tests were made on geforce gtx660, win7 x64, driver 372.70, Zoom Player 10, EVR and madVR.

[CarlosCaco]

Silly question...
I m using sanear audio renderer
On windows audio says my tv supports only 16 bit audio...
Should check only 16 bit in lav or should check all formats and let windows hang this?

[VictorLS]

Gleb Egorych,
are you tried my screenshot's settings of LAV Video Decoder? With this settings deinterlacing doing videocard without DXVA (same as CoreAVC Hardware Agressive Deinterlacing) and your sample must playing without artifacts.

[Gleb Egorych]

VictorLS, with your settings I still see artifacts, the checkbox has nothing to do with the way the decoder is initialized. If you look at source code of LAV Video you will see that on Win7 cuda decoder is always normally created via cudaVideoCreate_PreferDXVA. In LAV CUVID decoder there is a dependence on "Use HQ DXVA Processing" checkbox, if it is unchecked then cuda decoder is created via cudaVideoCreate_PreferCUVID.

Different behaviour on your system may mean that:
1. Something was changed in hardware (VP6 in GTX750 vs VP5 in GTX660).
2. Something was changed in software (driver) for 7xx.

[VictorLS]

Gleb Egorych
Now I understand - XP do not have DXVA2 so with CUVID your 1HD sample is playing without artifacts. In Win7x64 I have seen artifacts on your sample, but 1HD from SAT shows well now with both DXVA2 and CUVID in LAV Video Decoder.
So, we ask nevcairiel in near future add possibility to turn off cudaVideoCreate_PreferDXVA in LAV Video Decoder interface as in LAV CUVID Decoder did.

[Gleb Egorych]

In Win7x64 I have seen artifacts on your sample
That's what I am talking about. So GTX750 and GTX660 are the same in this case.
but 1HD from SAT shows well now with both DXVA2 and CUVID in LAV Video Decoder.
It's a floating problem, sometimes it appears, sometimes it disappears. Next time you watch 1HD or other channel it may rise again.
So, we ask nevcairiel in near future add possibility to turn off cudaVideoCreate_PreferDXVA in LAV Video Decoder interface as in LAV CUVID Decoder did.
Simple return to LAV CUVID decoder behaviour is not that good, because it will disable high quality deinterlacing. So we ask nevcairiel to try to fix it keeping HQ deinterlacing intact.

[VictorLS]

Gleb Egorych
Do you see difference between HQ deinterlacing on and off on HD? So, if you are know Russian language we can discuss it here http://forum.ixbt.com/topic.cgi?id=10:60902-43

[clsid]

The HQ deint option was removed on December 6th:
https://github.com/Nevcairiel/LAVFilters/commit/28315362f1a3327b5d12f1f9345fa4126854b33b

As you can see in the code, it currently is already disabled by default on XP. The only thing the option did is to prefer "DXVA mode" in the CUVID decoder. So 'fixing' decoding while also keeping deinterlacing quality is not possible. It is one or the other.

[Gleb Egorych]

VictorLS, search for "cheese slices deinterlacing", there are test videos with a how-to description. The difference is clearly visible.

So 'fixing' decoding while also keeping deinterlacing quality is not possible. It is one or the other.
Then how did CoreAVC make them work simultaneously? I understand that in current implementation the checkbox does not do the job so it is useless to simply return it.

[VictorLS]

Gleb Egorych
I have tried Slicies H.264 PAL 1080i-25.ts (from AVSForum) on XP (my BenQ FP241VW is 1920x1200 connected via DVI and I set 50Hz by nVIDIA's Control Panel). With check Enable Adaptive HW Deinterlacing it looks worse (more artifacts) then with unchecked in NVIDIA CUVID mode. On 7x64 I not see difference in LAV Video decode, but LAV CUVID Decoder shows worse with checked Use HQ ...
PS. I tried these Slicies in 2009 on AMD HD 3650 and nVIDIA GT220 - AMD was much better but IMHO it was only one preference over nVIDIA, but custom vertical frequency setup, colors, drivers, support and some other was made much better by nVIDIA.

[Gleb Egorych]

VictorLS, it's strange, for me HQ deinterlacing have an obvious advantage

HQ on:
http://s020.radikal.ru/i706/1609/81/d1c90941bb56.png [660 KB]

HQ off:
http://s56.radikal.ru/i154/1609/2d/01708b4ff68b.png [750 KB]

Done with LAV 0.67.

[andyvt]

I received a sample file that causes a crash in avutil-lav-55.dll. The same file plays via ffplay.exe. LMK if you need more info.

Thanks

Sample (http://babgvant.com/downloads/samples/test-crash.zip)

FYI - doesn't repro with github master

[VictorLS]

Gleb Egorych
With 0.68.1 (no git) in WinXP d1c90941bb56.png ~ as my HQoff and 01708b4ff68b.png as HQon )
Screenshots are impossible with VMR7w.
May be clsid explains us what happening.

[nevcairiel]

Windows XP has never supported HQ mode.

[VictorLS]

nevcairiel
Thanks for reply. But why I see difference between checked and unchecked Enable Adaptive HW Deinterlacing in XP.
Repeat: with checked worse.
And why LAV CUVID Decoder shows worse with checked Use HQ ... on Win7x64?

[nevcairiel]

It works how it works. Use it or don't.

[VictorLS]

nevcairiel
)
You are quite right and I agree in this case, but hope that in case of CUVID you'll implement HEVC 10bit support if it will be possible https://devtalk.nvidia.com/default/topic/965840
It is very very important for me. You are my last hope (if nVIDIA will not object) for smooth watching 3840x2160p50 (60) HEVC 10bit channels from SATs on WinXP with my new GTX750 GM206 - all necessary for me video including 4K HEVC 8bit is perfect with 0.68.1 - Thank you very much for your filters.
PS. I agree even on LAV CUVID Decoder i.e. 0.14 with correct 4K h265 10bit support.

[amayra]

is more than 4 months since last stable build how about last nightly is look stable enough for me

[AngelGraves13]

Has any progress been made on decoding the object based audio portion of Atmos and/or DTS:X?

What new features of bug fixes could we expect in the next stable version?

[huhn]

what's the point of decoding object based audio codecs?

you need a mixer/audio renderer that can handle this type of audio too.

[VictorLS]

max_verem from https://devtalk.nvidia.com/default/topic/965840 with philipl's help can hardware play 4K HEVC 10bit (as I understand with 8bit dithering) using CUVID on Linux already.
Developers of LAV Video Decoder, tell please is it possible, theoretically, to do the same with CUVID in LAV Video Decoder?
And one more question: why CUVID not work on Win10 totally?
PS. Despite of last question I need CUVID H265 10bit hardware decoding on WinXP and Win7x64.

[CruNcher]

@nevcairiel

what is that ?

every scene change in the Transformer (ATME Kyrion Encoder) sample causes a massive latency spike of 15 ms ?

http://demo-uhd3d.com/fiche.php?cat=uhd&id=22

Lav Video (15 ms latency spike every scene change)

http://i1.sendpic.org/t/mN/mNLogZaIXv5c2MPVg1xI6aPivu6.jpg (http://sendpic.org/view/1/i/31K6CWk9fjytKMFGwS1AH3pOvpQ.png)


Lentoid Decoder (no issue)

http://i1.sendpic.org/t/wn/wnl7fBDWIN8jAXP3hufbjReOFoX.jpg (http://sendpic.org/view/1/i/ypFKbwv5iq1Hv3iWDg8h63kf9Vl.png)


I Frame decoding performance issue ?

i guess it could match with this

Lav Video CPU

http://i1.sendpic.org/t/f4/f4Oppp66Wexv9MiIAESTetRV4w3.jpg (http://sendpic.org/view/1/i/gw5BsSf8m04CHQeCGnKR1PK2Yx2.png)


Lentoid CPU

http://i1.sendpic.org/t/t6/t6eICbkKQw85HNV8BkZiuh6MpCq.jpg (http://sendpic.org/view/1/i/x6OZDjn8t6JPWlHQfmRblcLFUch.png)

[AngelGraves13]

what's the point of decoding object based audio codecs?

you need a mixer/audio renderer that can handle this type of audio too.

The same reason we wanted to be able to decode DTS-HD MA...

To fully experience the audio track without compromises.

[huhn]

even if you can decode it you can't use it.

what is the compromises of bit streaming object based audio. every hardware has to bit stream it.

even if we could decode object based audio and send this decode data to your AVR it couldn't do anything with it.

[e-t172]

The same reason we wanted to be able to decode DTS-HD MA...

To fully experience the audio track without compromises.

DTS-HD MA is a "traditional" format. It decodes to a fixed number of channels, like TrueHD or FLAC. Object-based audio is a completely different paradigm entirely, which makes things harder. If the audio decoder outputs decoded object-based audio, the next step in the chain (the audio renderer) needs to know what do to with it. In practice, that probably means someone needs to write an audio renderer that is capable of correctly mixing and rendering object audio while taking end user setup into account (how many speakers and where they're located). That's a lot of work - way more work than would be required to support a traditional format.