LAV Filters - DirectShow Media Splitter and Decoders

[nevcairiel]

In the end, that means the correct way to mix a center channel (one sound source) into stereo (two sound sources) so that the sound pressure level remains the same is to send the center channel signal to both channels with a -3dB (x(1/sqrt(2)) attenuation, not -6dB (x0.5). This is quite counter-intuitive, so don't fall into the trap. ffdshow got this right (as seen by the 0.707 factor in the mixer matrix).
Thats whats being done in LAV as well, the coefficient for splitting one channel on two other channels is M_SQRT1_2, or 1/sqrt(2), or 0.707.
I looked up some reading material on digital mixing quite some time ago, even though i didn't really need to because the mixing is all done by libavresample (but at least i now understand why such things are done now :D)


Also, when downmixing 5.1 to stereo, I'm not sure applying an attenuation factor on the whole output because of clipping paranoia is such a good idea. AFAIK, according to Dolby guidelines, the content producer is supposed to check clipping doesn't happen even when the 5.1 track is downmixed to stereo, which makes sense because such downmixing is fully supported in Dolby specs. Attenuating after downmixing would cause some issues, like really low levels (most movies' average levels are already so low that most consumer equipment have barely enough gain for it to be comfortable, so don't make it worse) and high level differences between stereo and 5.1 content.
Sadly not all content is Dolby content.
Maybe such a mode as cyberbeing suggests isn't such a bad idea then, basically just clipping protection which only adds that attenuation factor when values exceed the -1.0 - 1.0 range.

[DragonQ]

Do you mean having the matrix with 1.0 values (rather than normalised ~0.3 values for 5.1 to 2.0) and then reducing volume whenever clipping is imminent?

[e-t172]

Sadly not all content is Dolby content.
Maybe such a mode as cyberbeing suggests isn't such a bad idea then, basically just clipping protection which only adds that attenuation factor when values exceed the -1.0 - 1.0 range.

Well, Dolby is the de facto standard, which means that if downmixing 5.1 content to stereo produces clipping, then the content producer is at fault, not the player which is just doing what 99% of players do.

In my opinion, both attenuation and clipping protection should be configurable and disabled by default. Especially clipping protection because, as useful as it is, it violates the principle of least surprise, especially when LAV is used as part of an encoding/production chain.

[nevcairiel]

In my opinion, both attenuation and clipping protection should be configurable and disabled by default. Especially clipping protection because, as useful as it is, it violates the principle of least surprise, especially when LAV is used as part of an encoding/production chain.

For one it would only ever be on when you turn on mixing, which is off by default. Assuming you manage to find your way to the option to turn on mixing, i would guess you managed to look at the other options that control the mixing process as well ;)
And at least when you output integer, where the values can never exceed their defined range, i think its preferable to actual clipping.

Turning on Stereo downmixing should be something thats relatively easy for a user, without users getting clipping on alot of content they watch because they didn't read the options, so either i default to the global attenuation (as its now still because of lack of options), or the clipping protection
I consider the average user to be more stupid then someone using LAV for content editing/production ;) The average user expects it to work, and is surprised if it doesn't :p
Again, all this only matters when they actually turned on mixing manually before, with mixing off (the default), the audio will not change at all to the current version.

I don't know what happens when you feed out-of-range float content to an encoder or the audio renderer, funny things may happen.

[kasper93]

https://dl.dropbox.com/u/16282309/LAV/00003.m2ts
Change to LPCM audio track, and playback is slowdown. Work fine with MPC internal splitter.

BTW. I have few VOB's in which LAV detect wrong time, but I tried to cut small part and LAV detect valid time, I guess it's ffmpeg bug, but works with MPC internal splitter.

[nevcairiel]

https://dl.dropbox.com/u/16282309/LAV/00003.m2ts
Change to LPCM audio track, and playback is slowdown. Work fine with MPC internal splitter.

I cannot reproduce any problems with that file.
Please try again with the build i posted earlier, http://files.1f0.de/lavf/LAVFilters-0.50.5-46-gd75bb42.zip

[wanezhiling]

http://forum.doom9.org/showpost.php?p=1425964&postcount=2
Q: DirectVobSub will not load with LAVFSplitter
A: Make sure to use the "DirectVobSub" filter when adding it to the preferred filters list, and NOT "DirectVobSub (auto-loading version)". The latter will NOT work.
This QA should be removed because current LAVF works fine with DirectVobSub (auto-loading version).:)

[Keiyakusha]

Btw yes, any of them works. Why autoload shouldn't be loaded if the only difference is that autoload version picks up external subs by default and non-autoload - not.

[kasper93]

I cannot reproduce any problems with that file.
Please try again with the build i posted earlier, http://files.1f0.de/lavf/LAVFilters-0.50.5-46-gd75bb42.zip

I found the problem, but I don't understand it. When I select "1: System Default" Audio Renderer in MPC-HC settings everything works fine. But when I select "2: Speakers..." it doesn't work. It's weird because it should be the same. Speakers are marked as default in windows settings. Anyway sorry for trouble, LAV is not an issue here :P

[ney2x]

Hi guys! Slightly off-topic... I just want to know what is the best settings of Reclock when using LAV Audio?
1. Direct Sound
2. WASAPI Exclusive
3. WaveOut

TIA.

[DragonQ]

WASAPI if you don't want to hear audio from any other application whilst watching your videos (this also ensures bit-perfect playback since it bypasses the Windows mixer). Otherwise, Direct Sound.

[Virtual_ManPL]

Is LAV Video Decoder with enabled "Use Stream Aspect Ratio" setting by default, handle files differently for some time or it's just my imagination,
because some of the files have 'broken ratio' like in old CoreAVC and now I need to disable this option.

[Sebastiii]

Hi,

I have a strange issue with this sample : https://dl.dropbox.com/u/10536084/lav/Sample/Somebody_That_I_Used_to_Know_Walk_off_the_Earth.mp4
It happen when ffdshow is set @ postprocess (it detect 25 fps, this clip is 23.976).
I have try different version of ffdshow and it's the same.
Try too LAV Suite 0.50.1 and it works.
It seems that with haali it's ok as splitter, so i don't know if it's splitter issue or lav video.
Without ffdshow as postprocess and if ffdshow the video is the decoder -> It's ok.
Thanks :)

[nevcairiel]

I have a strange issue with this sample : https://dl.dropbox.com/u/10536084/lav/Sample/Somebody_That_I_Used_to_Know_Walk_off_the_Earth.mp4

Why is that .webm file named .mp4? :p
The problem with that file is that it doesn't store a proper FPS in its headers, and the VP8 header also doesn't have one, so in the end we get something like 1 million fps.

I'll add a check to try to probe the correct FPS when such a wrong value is detected.

[Sebastiii]

I think it was named .mkv before lol
Thanks Nevcairiel :)

[clsid]

Is LAV Video Decoder with enabled "Use Stream Aspect Ratio" setting by default, handle files differently for some time or it's just my imagination,
because some of the files have 'broken ratio' like in old CoreAVC and now I need to disable this option.The problem is more apparent when using LAV Splitter instead of Haali, since Haali splitter modifies the stream with the container AR.
nevcairiel is considering making the option smarter, so that it automatically uses stream or container AR based on the reliability of the container format.

[Sebastiii]

Thanks Nevcairiel, latest push seems fix the issue :)

[Volfield]

Quick question to Nev. I have 3 subtitle in one mkv: eng (force), eng (default) and eng. Is it possible to set splitter that open file with eng not force and not default flag? If not can You add something like "*:eng|n" to set splitter to choose subtitle track without flag force or default?

[nevcairiel]

Quick question to Nev. I have 3 subtitle in one mkv: eng (force), eng (default) and eng. Is it possible to set splitter that open file with eng not force and not default flag? If not can You add something like "*:eng|n" to set splitter to choose subtitle track without flag force or default?

Its not yet possible, but i already have that on my list.
http://code.google.com/p/lavfilters/issues/detail?id=233

[tormento]

I cannot reproduce any problems with that file.
Please try again with the build i posted earlier, http://files.1f0.de/lavf/LAVFilters-0.50.5-46-gd75bb42.zip
Is it possible to get a x64 compile? ;)

[ryrynz]

Xhmikosr has it already done.

http://xhmikosr.1f0.de/lavfilters/

[kasper93]

Is it possible to get a x64 compile? ;)

http://forum.doom9.org/showpost.php?p=1580617&postcount=11417

[Keiyakusha]

About this mixing stuff. I don't really know how it works but for example if I have it disabled, then with stereo headphones setup i hear:
front left - left ear
front right - right ear
center - both ears, volume is relatively the same as one of the fronts
back left/right - i hear both in both ears but left is more loud in left ear, right - in right
lfe - I hear in both ears

if i turn ON mixing in LAV (and select stereo) i hear:
front left - left ear
front right - right ear
center - both ears
back left - left ear
back right - right ear
lfe - nothing

so in 1st case mixing is done by drivers. The question is, is this what i 'm supposed to hear? Actually I expected to hear from LAV something comparable to the job that my drivers doing...

[nevcairiel]

Thats a question of definition. Personally i don't think the back channels should bleed over into the wrong ear, but some matrix encoding from Dolby specifies it like that.
I already implemented options to choose between different matrix encoding types, "Normal" (now), and "Dolby Pro Logic" and "Dolby Pro Logic II". The two dolby modes have bleeding of the back channels into the other ear.

LFE is currently not being mixed because Dolby didn't recommend doing it, but i'll probably add a checkbox to do that.

In any case, if you're happy with the Windows Mixer (or the Drivers), no need to bother with it. ;)

[Keiyakusha]

In any case, if you're happy with the Windows Mixer (or the Drivers), no need to bother with it. ;)
I see. Thanks for explanation. I'm not very happy with it. Ideally I want to hear rear channels in one ear too like in case with LAV, however bass sounds really nice in my headphones so I want to have it too. I'm not sure If I can configure drivers to behave like that, so would be nice to have checkbox, if its not too much of a trouble.
EDIT: what i want is also the default foobar2000 behavior when downmixing to stereo is on.

[Nevilne]

What would be the closest setting to ffdshow default for 2.0 out?

[edwrap]

I see. Thanks for explanation. I'm not very happy with it. Ideally I want to hear rear channels in one ear too like in case with LAV, however bass sounds really nice in my headphones so I want to have it too. I'm not sure If I can configure drivers to behave like that, so would be nice to have checkbox, if its not too much of a trouble.
EDIT: what i want is also the default foobar2000 behavior when downmixing to stereo is on.

if you listen with headphones a lot, it might be worth it to look into headphone surround sound virtualization

for instance, foobar2000 has a dolby headphone plugin you can use to downconvert 5.1, but there are many other options as well (some driver software, cyberlink powerdvd's audio decoder, ffdshow, etc.)

there are also dolby headphone .dll's, but I've no idea if nev is interested in having lav audio directly use it nor do I know how much work that would entail

[Keiyakusha]

Yeah I know there is different things, but here I want just some minimum that will help me enjoy watching video. There is much more things nev planned to do, not only related to audio. And most of these are the features that I want too, so I don't think I can ask for more than adding LFE checkbox (^_^; )

[Pat357]

Nev,

Would it be a problem to implement the "Earwax effect" for headphones ?
I saw it's supported as a libav filter, so IINM, you only would need to put it in the chain and make a check-box for it.

I think it's a relative easy to implement thing... but as I read it in the docs, it seems to support only 44.1 kHz (Audio CD), so an extra resample step will be needed for common 48kHz audio.

What's your opinion about "Earwax" in Lavaudio ?

[mr.duck]

I have a TS sample file that LAVSplitter can't handle very well at all. Do you want it?

[Andy o]

Thats a question of definition. Personally i don't think the back channels should bleed over into the wrong ear, but some matrix encoding from Dolby specifies it like that.
I already implemented options to choose between different matrix encoding types, "Normal" (now), and "Dolby Pro Logic" and "Dolby Pro Logic II". The two dolby modes have bleeding of the back channels into the other ear.

LFE is currently not being mixed because Dolby didn't recommend doing it, but i'll probably add a checkbox to do that.


It's cause DPL(II)(IIx) is not supposed to be a downmixing algorithm for final stereo use, but a way to encode more channels into a stereo signal in order to be decoded later back into 5 channels. For headphones, Dolby Headphone can be used in conjunction with DPLII to hear the intended effect.

I guess encoding to DPLII is a nice feature to have, but I can't figure out a use for it, cause, for example, it helped with stereo analog audio content, or Gamecube and PS2 games, which didn't allow for in-game multichannel audio, but pretty much everybody now has a multichannel device on their computers. If it's a laptop or something, DPLII by itself is not meant for 2 speaker playback anyway, and if D. Headphone is available, you can just bypass the DPLII and play with DH directly the multichannel audio in the first place.

[Andy o]

I see. Thanks for explanation. I'm not very happy with it. Ideally I want to hear rear channels in one ear too like in case with LAV, however bass sounds really nice in my headphones so I want to have it too. I'm not sure If I can configure drivers to behave like that, so would be nice to have checkbox, if its not too much of a trouble.
EDIT: what i want is also the default foobar2000 behavior when downmixing to stereo is on.

Have you tried Dolby Headphone? It should be much better. In fact, it can be pretty impressive. Not every headphone virtualization will work though, in fact, most are crap and I understand completely the skepticism these produce. The only other competing one that works well is DTS Surround Sensation Headphone, but that's almost non-existent (only TMT offers it in the PC I think). DH devices like the Xonar U3 dongle can be found for around $40 or probably a bit less. Most Xonar cards and Realtek 889 (many Gigabyte mobos) offer it as well. If you can get ahold of the dolby headphone dll from Cyberlink (with any PowerDVD version that offers it, which is pretty much all since 7.3 or even earlier), you can try it out with foobar2000 and the Dolby Headphone wrapper (http://www.foobar2000.org/components/view/foo_dsp_dolbyhp) component. Try it with a 5.1 discrete test sample to see what it can do.

[dansrfe]

It's cause DPL(II)(IIx) is not supposed to be a downmixing algorithm for final stereo use, but a way to encode more channels into a stereo signal in order to be decoded later back into 5 channels.

I'm not sure I understand what you mean by this. DPL(IIx) works very well for final stereo use with my headphones and satellite speakers and it definitely sounds better than using straight up stereo downmixing. Moreover I don't think downmixing to stereo and decoding back to multichannel keeps the same quality as sending multiple channels straight obviously.

[Andy o]

If it sounds better to you it's incidental. Dolby Surround, Dolby Pro Logic, and DPLII are surround matrix algorithms. They're designed to be played in 4 (D. Surround) or 5-speaker setups. In the old analog tape days they were the only thing consumers had (available in a widely used popular format). It doesn't keep the same quality, but it was a sort of workaround for the technology available, and DPLII can be very effective. That's why it was used for GC and PS2 games, those weren't designed for discrete multichannel in-game audio.

http://en.wikipedia.org/wiki/Dolby_Pro_Logic
http://www.dolby.com/us/en/consumer/technology/home-theater/dolby-pro-logic-ii-details.html

The Dolby page is of course a bit of an exaggeration. For DPLII to work best (or even acceptably), it has to be specifically encoded, and either the "movie" or "game" modes used, not "music". It may not work well for just any stereo content.

[nevcairiel]

Those Dolby algorithms are designed in a way to sound good in stereo speakers as well, because obviously Dolby was aware that not everyone had a 5.1 setup when playing such tracks (especially in the early days), and if it sounded like crap in Stereo speakers, it wouldn't have taken off.

[Andy o]

They don't sound like crap, but they weren't designed to be listened in stereo either. Games for example, have a setting to enable/disable DPLII, which implies that, if available, listening to a dedicated stereo mix is preferable. IIRC DPLII applies some phase tricks, right? If so, couldn't there be some interference between the surround and front channels if played in the same channel? This would not be noticeable always.

I found this (http://www.dolby.com/uploadedFiles/zz-_Shared_Assets/English_PDFs/Professional/214_Mixing%20with%20Dolby%20Pro%20Logic%20II%20Technology.pdf). They briefly address this situation:

In addition to home theater viewers with matrix decoders, mixers must also account for viewers
listening to stereo and mono speaker setups. This includes people watching TV in their bedrooms,
dens, RVs, and other places where Dolby Pro Logic II decoders aren’t practical or desired.
Because of this, it is important to listen to the mix in both stereo and mono to make sure the mix is
acceptable for these setups.

So, the difference with commercial, professionally encoded, DPLII content, and encoding on-the-fly, is that consumer DPLII content should have been double-checked so it sounds "acceptable" in stereo/mono.

[jmone]

Having options is great, but .... (not making any friends in this thread I guess), I'd argue that a decoders job is to... decode... and it should be the Media Players job to worry about how to handle the decoded audio (mixing, room correction, dsp etc etc). Then again I'm an MC user so we have had these functions working with LAV for awhile now. If your player does a poor job of this then it may be time to upgrade to a better one :)

[Virtual_ManPL]

The problem is more apparent when using LAV Splitter instead of Haali, since Haali splitter modifies the stream with the container AR.
nevcairiel is considering making the option smarter, so that it automatically uses stream or container AR based on the reliability of the container format.

In my case splitter didn't make any difference. But I tested older versions of LAV Filters to see what's the cause and I saw that bug occurred there too. So I reported it on bugtracker, because with MPC-HC internal decoder it works propertly.
bug #241 - Files with different DAR & original DAR have wrong aspect ratio with LAV Video Decoder (https://code.google.com/p/lavfilters/issues/detail?id=241)

[nevcairiel]

"Original display aspect ratio" is the AR from the AVC stream, so if you play that file with "Use Stream Aspect Ratio" enabled, it should come out as 3:2, or 720x480 exactly.
Only if you turn that option off, the Container AR is used, which Media Info calls "Display aspect ratio", and is 16:9.

If anything else is happening, that would be a bug, but if you get 3:2 with the file you linked, its working as intended. The filehoster you used is blocked by my work firewall, so i cannot download it right now.
If MPC-HC outputs the file with 16:9 despite the "Read AR from stream" option enabled, then that option may just be bugged in their decoder.

[Virtual_ManPL]

I updated bug info, added screenshots and multiupload link
Seems like you're right. LAV works properly and MPC-HC is bugged.
Odd because in MPC-HC it looks more natural.

[nevcairiel]

Its quite common that people don't properly flag the AR in the bitstream, and instead only modify the container, so i guess 16:9 is what its supposed to be played at, but how would LAV know that?

Like clsid explained, i plan to add an option for the Stream AR setting which does not use the Stream AR on containers which commonly have this problem, like MKV, MP4 and AVI (actually, thats it, can't think of more important containers with that problem).
I'll probably add it for the next version even.

[DragonQ]

That'd be really useful! I need "use stream aspect ratio" on for some of my older 1440x1080i recordings to properly playback as 16:9 (they are TS files). However, I also have some MKV files where the stream has an incorrect aspect ratio and I've had to fix this using the container aspect ratio (almost always using MKVMerge). In this situation, I need to untick "use stream aspect ratio" to make the container aspect ratio have priority.

If it used the container aspect ratio where possible but stream aspect ratio where it wasn't possible, that should be perfect, if I've understood correctly... :)

[nevcairiel]

It would be as simple as only using the container AR on containers where its more reliable, like MKV/MP4, and stick to Stream AR for MPEG-TS and others.

[Lincoln Burrows]

Folks,
I asked in the old thread:

http://i.imgur.com/rQnsg.png

Here's the thing: I noticed it's necessary to change that "output mode" to 25/30 instead of 50/60 while watching 1080i contents (saved in Matroska (lossless) or not), otherwise we face audio sync problems. Is that change necessary? Once I changed, it seemed like the playback was fine.

An example of content using 1080i would be a TV show aired in UK, called TORCHWOOD, in Blu-ray. Or the 1st season from CSI: Los Angeles (2000).

It should be noted I don't have to change anything for any other content. Only 1080i.

Another issue I am having is this one:

Have you guys had any trouble playing WMV videos in MPC-HC?

I have a Q9450, 2 GB/RAM (I know, insufficient memory, but I can still use normally), Windows 7/32bit. I never had problems playing HD videos until... now. I tried a WMV file with 3 GB and 30 minutes, full-HD (1920x1080), and... my PC was so slow, I mean, so slow, that I had to force a reset. In the second time I had to use control+alt+del, and I saw it wasted at least 1 GB/RAM to play that file. This thing was soooooooooooooooooooooooooooooooooooooooo slow that it was unbelievable!

However, I tried using VLC to play that same file and it worked... no slowdowns, no problems...But I believe it might be related to MPC-HC, since the video worked in VLC. Any clues on what happened?

P.S. I also have a 8800 GTS 512 MB video card.

[betaking]

Folks,
I asked in the old thread:

http://i.imgur.com/rQnsg.png

Here's the thing: I noticed it's necessary to change that "output mode" to 25/30 instead of 50/60 while watching 1080i contents (saved in Matroska (lossless) or not), otherwise we face audio sync problems. Is that change necessary? Once I changed, it seemed like the playback was fine.

An example of content using 1080i would be a TV show aired in UK, called TORCHWOOD, in Blu-ray. Or the 1st season from CSI: Los Angeles (2000).

It should be noted I don't have to change anything for any other content. Only 1080i.

Another issue I am having is this one:

But I believe it might be related to MPC-HC, since the video worked in VLC. Any clues on what happened?

P.S. I also have a 8800 GTS 512 MB video card.

8800 GTS VP2 not support VC-1 cuvid!

[CiNcH]

I have been using the ffdshow audio mixer so far and I have always been asking myself whether it is possible for the filter to auto-detect the output speaker configuration. I am often switching between the stereo speakers of my TV (via HDMI audio device) and the analog surround speaker system (via my 7.1 soundcard configured for 5.1 output) when I am watching with my projector.

[DragonQ]

Here's the thing: I noticed it's necessary to change that "output mode" to 25/30 instead of 50/60 while watching 1080i contents (saved in Matroska (lossless) or not), otherwise we face audio sync problems. Is that change necessary? Once I changed, it seemed like the playback was fine.

An example of content using 1080i would be a TV show aired in UK, called TORCHWOOD, in Blu-ray. Or the 1st season from CSI: Los Angeles (2000).

It should be noted I don't have to change anything for any other content. Only 1080i.
Torchwood is likely 1080p (in a 1080i "wrapper"), so you'd see no difference between 25p and 50p deinterlacing. Try some proper interlaced content like any sport.

It could be that your system can't handle deinterlacing to 50p but I kinda doubt it since I have a GTS250 (which is a re-brand of that card) in my desktop and it works fine for me. I'm using CPU decoding though, with DXVA2 deinterlacing. Dunno if the card is good enough for CUVID/DXVA2 decoding and DXVA2 deinterlacing.

[Pat357]

It would be as simple as only using the container AR on containers where its more reliable, like MKV/MP4, and stick to Stream AR for MPEG-TS and others.

Where are .MOV's classified ? Same as MP4 ?
You find these a lot for trailers, youtube,...
Even my Canon outputs 1080p (23.976, 24, 25 or 29.97) H264 in a MOV container.

[nevcairiel]

MOV is MP4 for me, all the same shit :p

[kalston]

Have you tried Dolby Headphone? It should be much better. In fact, it can be pretty impressive. Not every headphone virtualization will work though, in fact, most are crap and I understand completely the skepticism these produce. The only other competing one that works well is DTS Surround Sensation Headphone, but that's almost non-existent (only TMT offers it in the PC I think). DH devices like the Xonar U3 dongle can be found for around $40 or probably a bit less. Most Xonar cards and Realtek 889 (many Gigabyte mobos) offer it as well. If you can get ahold of the dolby headphone dll from Cyberlink (with any PowerDVD version that offers it, which is pretty much all since 7.3 or even earlier), you can try it out with foobar2000 and the Dolby Headphone wrapper (http://www.foobar2000.org/components/view/foo_dsp_dolbyhp) component. Try it with a 5.1 discrete test sample to see what it can do.

The free Isone Surround (http://www.toneboosters.com/tb-isonesurround/) (easy to use in JRiver MC17 or foobar2000, a bit tricky in ffdshow but still doable) is IMO much better than Dolby Headphones or the DTS thing in TMT. It's actually the only headphone virtualization I've heard that sounds really good to me.