LAV Filters - DirectShow Media Splitter and Decoders

[cremor]

Ok, no problem then. Didn't know that the splitter can't read that values.

[Sebastiii]

So I've got mpc-hc 32bit, LAV filters 32bit and I put the dtsdecoder.dll (the same one that works fine with eac3to) in the LAV filters folder, but it still doesn't seem to decode dts-hd.

I downloaded dts-hd samples here http://www.demo-world.eu/trailers/high-definition-trailers.php :
"Orchestra (Short) (Lossless) 1 57,6MB 0:14 DTS-HD HR 7.1" and "SFX (Long) (Lossless) 1 93,6MB 0:19 DTS-HD MA 7.1"

No luck so far. I've tried disabling the dca decoder in LAV audio but it didn't help.

For me i must set dca and if i'm not wrong arcsoft dll is only needed if you don't bitstream to AVR.
For it's working that way, but DTS-HR didn't work in bitstream and seems related to my AVR even if with PDVD it's OK (but not TMT5).
It's quite odd lol.

[jmone]

With LAV CUVID you mean with the real LAV CUVID, or with LAV Video in CUVID mode?
Does it only happen in software mode?

Can you try another renderer?

It does sound like internally the resolution is correct, but madVR thinks its something else - which accounts for the chroma offset. But that doesn't make any sense, how would that happen. The internal resolution is directly put into the media type.

OK - Did some more testing

Chroma offset occur if:
- LAVVideo in SW Mode
- MS Video decoder

Chroma offset does not occur if any of the following is changed:
- LAVVideo in CUVID Mode
- LAVVIdeo in SW Mode if using a 4:2:2 output (eg disabled both 4:2:0 options...the movie res is still incorrect in madVR but it works fine)
- FFDSHOW Video Decoder (the version used by JR)
- EVR

I'm also having trouble uploading a file as soon as I create a sample (or even copy it to another location) the issues goes away....and then I noticed another pattern. These files were created from Music DVD's with DVD Decryptor to break them by chapters (eg one file per song). It is only the First file in each set created this way that has the issue.

It is all very odd but happens on two PC's.

[nevcairiel]

Since its apparently not isolated to LAV Video only, i would suggest to also report that to madshi. A sample to reproduce it will be required in some shape or form, however. Don't remux it, just cut the first few MBs off and see if it still reproduces. (Try DGSplit or something)
Is that actually a new issue since he added deinterlacing?

PS:
4:2:2 doesn't show the issue because its a packed format - there is no separate chroma plane.

Edit:
Wait, i just read - copying the file around changes the behaviour? Your system seems seriously screwed up. :p

[pacemaker1000]

What do you have connected to the HDMI? The speaker configurations listed there comes from info (EDID) sent back from the connected device.

This is now OT for this thread (since your problem has nothing to do with LAV) so you should continue this matter somewhere else. There are threads about EDID on AVS Forums and other places.

sorry but without the advice on here, i thought it was a lav setup problem, i wouldnt have realised the problem was the speaker vonfig

thanks especially to SamuriHL for helping me sort it

[jmone]

Since its apparently not isolated to LAV Video only, i would suggest to also report that to madshi. A sample to reproduce it will be required in some shape or form, however. Don't remux it, just cut the first few MBs off and see if it still reproduces. (Try DGSplit or something)
Is that actually a new issue since he added deinterlacing?

PS:
4:2:2 doesn't show the issue because its a packed format - there is no separate chroma plane.

Edit:
Wait, i just read - copying the file around changes the behavior? Your system seems seriously screwed up. :p

Found the "X" factor...It is the crop edges command in MC (hence why it was not showing up in my sample...until I reapplied crop edges!). If some of these files are tagged with "crop edges" you get the chroma shift as described above with certain combinations of decoders and madvr.

Given it is some three way thing, I'll post the sample in the MC Beta Forum.

[jmone]

EDIT - posted on the MC Beta forum and let madshi know as well.

[Plutotype]

Hi Nev,
Recently I have reported a weird issue at madVR forum. Instable decoder queue, which drops from 11-12/12 to 1-12/12 or lower causes the presentation queue drop also to 0-8/8, which leads to frame drops. I sent madshi couple of logs and he commented this as follows:

"According to the log, playback is running just fine. Then all of a sudden the decoder stops sending frames to madVR for a short while. Of course audio continues to play in the meanwhile. So once the decoder catches up again, madVR has no other choice than to drop a couple of frames to make sure that audio and video stay in sync. Now what you need to find out is why the decoder sometimes freezes for a short while. This is outside of madVR's control, as far as I can say. It could be caused by a bug in the decoder, or by your PC being very busy for a short while, so that no CPU time is left for the decoder at that time. My first suggestion would be to try a different decoder, just to make sure it's not the decoder. If it's not the decoder (and I guess it's not) then you should run some profiling tool which records CPU consumption during video playback. Then when you get these frame drops, stop playback and check whether the profiling tool has recorded 100% CPU consumption shortly before.
The decoder queue is supposed to stay nearly full (at least 8/12) at all times."

One minute of the tested file can be found here: http://www.mediafire.com/download.php?1rja6beiptppzq8

What makes the decoder to constantly decode frames and delivering full queue to the renderer? What causes that some frames decoding is delayed and the decoding queue goes from 12-11-10-...3-2-1 to 0 and causes presentation issues? Madshi hinted to my CPU utilisation, but with softcubic 80, spline 4, spline 4 it is 10percent on my i7980X:). CPU is not downclocking, GPU is not downclocking.

Thanks for your opinion.
Pluto

PS: The best results so far ( 0 dropped frames ) I got with Microsoft DTV DVD Video Decoder ( decoder queue is constantly 11-12/12 ), but need to test full movie with that yet and compare results with other decoders. Other decoders behave more unstable - occasional small drops in frame decoding increases the risk of presentation issues.

[pankov]

Nev,
did I miss your answer or you missed my question there ... or you simply prefer not to answer it right now?
I understand - you are personally ignoring me.
...
Just kidding
;)

Now that madVR supports DXVA deinterlacing I'll no longer bother you with that question.

[SamuriHL]

ROFLMAO :) Honestly speaking, to me it makes sense to have the deinterlacing done by the renderer. Truly awesome. Now, that being said, there may be some cases where yadif in the decoder might make sense for some people. I'm definitely covered on both machines now. Absolutely wonderful!

[Pix]

I noticed a problem. Don't know if it has been mentioned before.
.mkv files that have video/audio(don't know which of these causes the problem) stream set to "Default track flag:" -> 'Yes' during muxing(mkvmerge) with chapters aren't seekable/using seekbar is insanely difficult.
When I want to jump to the 2nd/3rd/etc chapter the whole video freezes. After 30-180 seconds the audio starts but the picture stays frozen and if I click on the same spot on the seekbar, sometimes, video continues to play.
If I try to glide the button on the seekbar - doesn't work. The video freezes, button is positioned on a new spot but the video isn't playing.

I did make sure that it wasn't something else. I changed the renderers, decoders, tried several different combinations.
When I changed the Splitter the problem disappeared. I changed the "Default track flag:" -> 'default' during muxing and the problem also disappeared.
Do take in mind that files with "Default track flag:" -> 'Yes' that have no chapters also don't have the problem.

Any settings I might be using that caused the problem?

[madshi]

there may be some cases where yadif in the decoder might make sense for some people.
Which cases? For people who want to post process with AviSynth? If they want to do that, they have to use the ffdshow raw video processor (or ffdshow video decoder), anyway, cause that's the only way to access AviSynth in DirectShow. So why then not using ffdshow's yadif deinterlacer?

There's another problem: yadif is a pure video mode deinterlacer, it's not meant to be used for movies. So it's only half of a deinterlacer solution. For movie content you need IVTC. When talking about IVTC, there are a multitude of cadences. IVTCing just NTSC's 3:2 is pretty easy. But what about PAL's 2:2? Much more difficult, because there are no duplicate fields. Then there are various Anime cadences. Then there are bad edits. Finally, broadcasts sometimes have mixed video + movie content. With such content you need to deinterlace some parts of the frame with yadif and some with IVTC. In order to make that possible, you need an automatic per pixel video vs movie type detection, which is extremely hard to get right, especially with PAL content. Deinterlacing is a very very complex and difficult topic.

Of course I don't want to stop nevcairiel from adding deinterlacing to LAV Video Decoder, I'm just wondering whether it's really worth his time, considering the complexity involved in creating a "complete" solution.

DXVA deinterlacing has the luxury of having most of these problems already covered. Ok, I'm not sure how good all of these ATI/NVidia algorithms are, but in theory they should do all what I described above, including mixed video/movie content, minus decimation, however.

[SamuriHL]

Please don't misunderstand me, madshi, I personally don't know of any situations where yadif would be preferred. IMO, hardware deinterlacing always beats software deinterlacing. I made that statement because there could be cases that I'm not aware of that MIGHT make sense. I can't think of any. :) Personally, I'm EXTREMELY pleased. In fact, I want to test using CUVID for decoding and madVR's DXVA for deinterlacing on my nVidia machine when I get a chance. I don't know if it'll make any difference at all, but, I do believe that deinterlacing belongs in the renderer. That's a personal opinion, so, if others don't agree, I'm ok with that. :) Nonetheless, I'm completely covered now on both machines. LAV Splitter, LAV Audio with DTS-HD MA decoding, LAV Video (with CUVID on my nVidia box), and now DXVA deinterlacing for hardware deinterlacing on my AMD machine. Ohhhhhhhhhhhhhh yeaaaaaaaaaaaaa. :) Thank you to both you and Nev!

[madshi]

Well, if people do want to post process with AviSynth, they will have to deinterlace outside of the renderer, because most post processing algorithms (e.g. scaling) require progressive video. So there are arguments for offering deinterlacing in a DirectShow filter other than the renderer. Maybe in the long run all post processing can be moved to pixel shaders? Not sure...

[SamuriHL]

That would be nice. I don't do any post processing at all on my machines. My requirements are very simple...decode the video and audio without messing around with it unless it's necessary, render both at the highest possible quality, and if necessary, using a ReClock like functionality to keep everything smooth. I have all that now with hardware deinterlacing. Absolutely phenomenal IMO. :) The comment I made in reference to having a software deinterlacer maybe being preferable was exactly for the people who have far higher requirements and more complex environments than me. I just don't know where they might need software deinterlacing. I don't, so, I'm now happy. And everything works PERFECTLY in my favorite player. I can't ask for better. :)

[asasadad_1]

What exactly is supposed to be the problem?
Seems to play just fine.

Part of the sound has disappeared,Sound and image are not synchronized.anyway,if ffdshow audio decoder being used,you will find the clear difference.

[Aleksoid1978]

Hi nevcairiel.
A see that you set Interlace flag after parse header and if video is Interlace.
But - VC-1 stream on HD-DVD always have interlace flag set to true, even if is progressive.
Link to example - http://www.techpowerup.com/downloads/530/HD-DVD_Demo_1080p_VC-1_DDPlus_5.1.html

[STaRGaZeR]

Of course I don't want to stop nevcairiel from adding deinterlacing to LAV Video Decoder, I'm just wondering whether it's really worth his time, considering the complexity involved in creating a "complete" solution.

Yeah, I'm sure that's the case here :sly:

[Mangix]

is there any way to downmix 6 hannels to 2? i ask because my TV accepts only 2 channel PCM and 6 channel ac3.

[nevcairiel]

Hi nevcairiel.
A see that you set Interlace flag after parse header and if video is Interlace.
But - VC-1 stream on HD-DVD always have interlace flag set to true, even if is progressive.
Link to example - http://www.techpowerup.com/downloads/530/HD-DVD_Demo_1080p_VC-1_DDPlus_5.1.html

So?
Thats not avoidable, and doesn't really hurt.
These flags are only not set if i'm 100% sure that the content is progressive and nothing else - and with VC-1 on EVO you just cannot be sure.


I understand - you are personally ignoring me.
...
Just kidding
;)

Now that madVR supports DXVA deinterlacing I'll no longer bother you with that question.
Didn't i post some test versions here? Anyhow, it did work fine, at least for NVIDIA, on AMD/ATI it was too slow. madVR has the advantage that it doesn't need to download the image from the GPU anymore, it can just keep it on there. :)
But as you noted, its no longer required to have it.


Which cases? For people who want to post process with AviSynth? If they want to do that, they have to use the ffdshow raw video processor (or ffdshow video decoder), anyway, cause that's the only way to access AviSynth in DirectShow. So why then not using ffdshow's yadif deinterlacer?
I can think of some cases, where you want to transcode and need progressive video, but don't want to go all out on adding ffdshow as a post-processor for simplicity - or with a renderer that just doesn't do deinterlacing still (Haali, Overlay, etc). I'm still considering implementing it, its really not much work, but i guess the time table got pushed back a bit.

[madshi]

Yeah, I'm sure that's the case here :sly:
Huh ? :confused:

[nevcairiel]

Part of the sound has disappeared,Sound and image are not synchronized.anyway,if ffdshow audio decoder being used,you will find the clear difference.

I did not notice any problems with the file, seems to play just fine.

[PeQuE]

Hi nevcariel,

I've uploaded a video (in fact, recorded mpeg2 TV .ts) that Lav Video is failing at deinterlacing from time to time.

http://dl.dropbox.com/u/15222696/MP_20111005_18-05_Bar%C3%A7a%20TV_Temps%20de%20joc.%20La%20Lliga.ts

As a side comment, Win7 MS DVB included codec for MPEG2 is also failing the same way, while Cyberlink (9 & 11) Video/SP Decoder is doing it well.

Hope you can take a look.

Thanks a lot!

Hi nevcariel!

Did you finally had the time to take a look at that video?

Thanks a lot.

[asasadad_1]

I did not notice any problems with the file, seems to play just fine.

I have found what caused the difference between ffdshow audio decoder and lav audio decoder.
"Mixer→output speakers configuration→2/0/0-stereo" is my default ffdshow audio decoder configuration,it works fine with that sample.
If I unckeck "Mixer" in ffdshow audio decoder configuration,it has the same performance with LAV Audio Decoder(part of the audio has disappeared,audio and video are not synchronized).
Thank you nev for looked into that sample:thanks:.

[andyvt]

Which cases?

Transcoding.

[SamuriHL]

See I knew someone somewhere would have a use case where non-rendering deinterlacing might be preferred. :D

[nevcairiel]

LAV Filters 0.39

LAV Splitter
- Fixed a bug that caused MPEG-2 and VC-1 parsers to produce wrong timestamps after a seek
- Fixed playback of raw .h264 files

LAV Video
- NV12 is now the preferred YUV 4:2:0 output format
- Fixed handling of the "As Input" RGB output setting
- Improved VP6 cropping
- Don't report completely progressive movies as interlaced to the renderer
- Fixed playback of some broken MPEG-2 MKV muxes
- RGB will now by default be offered as Bottom-Up instead of Top-Down for improved compatibility


Download: Installer (both x86/x64) (http://files.1f0.de/lavf/LAVFilters-0.39.exe) -- Zips: 32-bit (http://files.1f0.de/lavf/LAVFilters-0.39.zip) & 64-bit (http://files.1f0.de/lavf/LAVFilters-0.39-x64.zip)

Before anyone asks again "Why NV12?" - its quite simple: Hardware post-processing usually requires the video to be in NV12, and especially on ATI only NV12 is accepted. Since there are no downsides to using NV12, it just makes more sense.

Another quick note, as a consequence of now trying to figure out if a movie is interlaced or not, the option to disable interlaced flagging has been removed.
Also, if you find a movie that was deinterlaced properly before, but now fails, please report those.

Have fun.

[fastplayer]

You can pretty much release after every commit! :p
The trunk builds have been very stable so far. *knock on wood*
:thanks:

[ikarad]

Which cases? For people who want to post process with AviSynth? If they want to do that, they have to use the ffdshow raw video processor (or ffdshow video decoder), anyway, cause that's the only way to access AviSynth in DirectShow. So why then not using ffdshow's yadif deinterlacer?

There's another problem: yadif is a pure video mode deinterlacer, it's not meant to be used for movies. So it's only half of a deinterlacer solution. For movie content you need IVTC. When talking about IVTC, there are a multitude of cadences. IVTCing just NTSC's 3:2 is pretty easy. But what about PAL's 2:2? Much more difficult, because there are no duplicate fields. Then there are various Anime cadences. Then there are bad edits. Finally, broadcasts sometimes have mixed video + movie content. With such content you need to deinterlace some parts of the frame with yadif and some with IVTC. In order to make that possible, you need an automatic per pixel video vs movie type detection, which is extremely hard to get right, especially with PAL content. Deinterlacing is a very very complex and difficult topic.

Of course I don't want to stop nevcairiel from adding deinterlacing to LAV Video Decoder, I'm just wondering whether it's really worth his time, considering the complexity involved in creating a "complete" solution.

DXVA deinterlacing has the luxury of having most of these problems already covered. Ok, I'm not sure how good all of these ATI/NVidia algorithms are, but in theory they should do all what I described above, including mixed video/movie content, minus decimation, however.
DXVA don't make IVTC deinterlacing.

I try since two years with mpc-hc or ffdshow and it doesn't work. (I have gtx285)

To use IVTC the only thing that works is the use of avisynth IVTC filter like decomb or TIVTC.

Do you expect to add IVTC filter in madvr?

[Dogway]

Now I don't know whether you missed my suggestion, you read it and added it to the todo list, or you read it but dismissed it

[nevcairiel]

Now I don't know whether you missed my suggestion, you read it and added it to the todo list, or you read it but dismissed it

The one on audio delay?
I didn't really know what you wanted. Its already possible to change it on the fly. Change the option, new delay applies (after clicking apply).

If you want even better "on the fly", use MPC-HCs delay option, you can just press the numblock +/- keys to change it.

[cremor]

"Orchestra (Long) (Lossless)" and "Orchestra (Short) (Lossless)" from here (http://www.demo-world.eu/trailers/high-definition-trailers.php) are reported as interlaced during playback (and even on that website) but they drop and delay a lot of frames using deinterlacing (madVR 0.78) and look a lot better without deinterlacing, so I assume they are progressive.
Anything the splitter/decoder can do here or were they just encoded wrong?

Btw, nothing new in 0.39, is the same in 0.38.

[Dogway]

Yes, you are right it works. It wasn't working for me because I had ffdshow audio processor after lav audio, I have that principally for audio downmix and volume booster. I by default disable audio switcher module as I can already do audio delay on lav audio. Thanks for the help.

[madshi]

DXVA don't make IVTC deinterlacing.
DXVA does do IVTC, kind of, but with limitations. There are 2 key problems with DXVA IVTC, from what I can see:

(1) The current VMR, EVR and madVR implementations of DXVA deinterlacing "blindly" trust the video bitstream flags. That works well for some DVDs/broadcasts, but not for all. The better solution would be to distrust the flags. This is currently not done by any renderer using DXVA, as far as I can say. I might implement that in madVR in a future version.

(2) There's no support for decimation in DXVA1/2, so the output will be 60p, not 24p. I've seen that there may be some sort of support for frame rate changes and maybe decimation in DXVA-HD, but I'm not 100% sure.

Anyway, this is probably a bit out of topic here, so let's not let get this into a lengthy discussion.

[CruNcher]

Wait wait

how can this be then http://www.mediafire.com/download.php?d0bg6khk2lk8bjl ???

[Boo]

I had some trouble playing this file: http://www.mediafire.com/?t52j52cp1v4gw96 (xvid AVI with max_bframes 0)
with MPC-HC - LAV Splitter, Video, Audio and madVR.

I can play the file with other renderers (Haali, EVR etc.)

But I wanted to be able to use madVR with Lav Filters so I started trying different settings.

I narrowed it down to the 'Threads' setting in the LAV Video Decoder.
I can play the file If I set threads to '1' aything else is a no go.

Am I the only one with this strange behaviour?

[mindbomb]

hi nev.
My friend claims that whenever he muxes something large with mkvmerge, the resulting mkv file is unseekable with lav splitter.
It appears there is nothing remarkable about his mkvmerge settings, but I may be wrong about that.
He muxes with an mkvmerge build for linux.

Is there anything in lav splitters mkv handling that could result in problems with files over 4 gigs?
I know this isn't a lot to go on, but it is hard to get samples for very large files.

Haali and mpc matroska filter have no issues with these files.

[Shark007]

10bit decoding seems to be broken with the .39 release.

[nevcairiel]

10bit decoding seems to be broken with the .39 release.

I cannot reproduce any issues with the 2 10-bit samples i happen to have on my laptop with me.

[golagoda]

I've been able to play 10-bit video perfectly fine also, and have heaps of 10-bit encodes which I just tested and all worked fine. Must be something wrong on your end Shark.

[pankov]

nev,
sorry to bring up such an old problem but could you take a look at the problem we discussed a few months ago here
http://forum.doom9.org/showpost.php?p=1498358&postcount=2509
and here
http://forum.doom9.org/showpost.php?p=1498551&postcount=2546
and finally here
http://forum.doom9.org/showpost.php?p=1498701&postcount=2570
you say you'll look into it but I can't find if you fixed it. I tried testing with older version but strangely I can't make them work on my PC !?!? (very strange but I'll dig into it later).
The real problem is that the latest versions (0.38 and 0.39 at least) still (or again) don't work on these samples.
(btw if you didn't store the PM with the FTP details I can send them again)

[nevcairiel]

Are those the fashion week videos? They used to work just fine, did that change?
I can test them when i get back home.

[pankov]

yes,
they are.
Currently they don't work .... but I'm not 100% sure they were before.
If you need me I'll try to test with older versions and report back but this can happen tonight when I get back home.

[nevcairiel]

I may actually have found the reason for it, some sneaky changes in ffmpeg might've broken my own sneaky changes to allow seeking in such files.
I'll check it out in detail later. It probably broke with 0.38.

[pankov]

You are right - 0.37 works fine
:)
sorry for doubting that you've actually fixed it before

[Shark007]

I cannot reproduce any issues with the 2 10-bit samples i happen to have on my laptop with me.

OK - this is what happens when I give bug reports just before going to bed.
I can confirm, the problem is/was with my setup and NOT with the LAV filters.

Thanks for checking into this nev, and especially thanks, for your time.

[TheShadowRunner]

Hi nev, just wanted to thank you again for finishing the job on the VP6 cropping issue.
With 0.39's Video Decoder, no more issues whatsoever regardless if I use LAV splitter or Gabest's FLVSplitter.

[nevcairiel]

You are right - 0.37 works fine
:)
sorry for doubting that you've actually fixed it before

For the record, its fixed again. :p

Hi nev, just wanted to thank you again for finishing the job on the VP6 cropping issue.
With 0.39's Video Decoder, no more issues whatsoever regardless if I use LAV splitter or Gabest's FLVSplitter.

Good to know its working fine now.

[PeQuE]

Hi nevcariel,

I've uploaded a video (in fact, recorded mpeg2 TV .ts) that Lav Video is failing at deinterlacing from time to time.

http://dl.dropbox.com/u/15222696/MP_20111005_18-05_Bar%C3%A7a%20TV_Temps%20de%20joc.%20La%20Lliga.ts

As a side comment, Win7 MS DVB included codec for MPEG2 is also failing the same way, while Cyberlink (9 & 11) Video/SP Decoder is doing it well.

Hope you can take a look.

Thanks a lot!

Sorry to bother you again in few days... but... Did you give a try to that video when you came back from vacation? It's the only tiny bug preventing me to finally use Lav Video as all-terrain decoder (live TV included)...

Thanks a lot nevcariel.

[nevcairiel]

Sorry to bother you again in few days... but... Did you give a try to that video when you came back from vacation? It's the only tiny bug preventing me to finally use Lav Video as all-terrain decoder (live TV included)...

Thanks a lot nevcariel.

I answered you before on that video. Its just a broken encode. Half of the frames are marked progressive - or at least seem to be.
I still have some small things to check, but all in all it doesn't look good.

I could try to offer "aggressive" deinterlacing, which means if MPEG2 is marked as interlaced, it would always treat all frames as interlaced.