LAV Filters - DirectShow Media Splitter and Decoders

[nevcairiel]

In the options menu of the Audio Decoder, I've noticed a setting titled “Auto A/V Sync correction”.

What exactly does this setting do? Does it tie into LAV Video to ensure both streams start playing at the exact same time?

I really should just remove that option, turning it off does usually never help.

Anyhow, what LAV Audio does is completely re-compute the timestamps of the audio based on the resulting PCM samples. This is done mostly to get rid of jittery timestamps in source files, providing a 100% smooth and jitter free PCM signal.

Now what that option does is watch the incoming timestamps from the source, and make sure they don't drift apart. If a file is undamaged, that should never happen, but if you watch a recording from DVB (or live DVB), its certainly possible that some data was lost in between, and there is a timestamp gap. Thats where the option would come in and re-sync the data.

Like i said in the beginning, thats how it should always work, and having an option to disable it doesn't really make all that much sense. It was mostly for debugging at first to make sure the new timestamp code works properly. :)

In the splitter stream switch menu all the info is lower case. Is there a reason for this? Could it be changed to proper mixed case?
The reason is that i like it that way. :D
Also, decoder names in ffmpeg are all lower case, so to make it look somewhat saner, i just lowercased everything.

Also level isn't indicated for MPEG-2 only the profile, but it is for H.264 and VC-1.

It could probably be added if the MPEG2 decoder outputs that information in ffmpeg, i'll look.

And track titles in MP4 arn't displayed. This would be nice to have.

I just display whatever title ffmpeg provides in the stream metadata.
Sample file with stream titles, and something that actually shows them? :) I don't have all that many mp4 files..

[Xaurus]

nev,

I think the main reason for me to still stick with ffdshow audio is because of the volume adjustments that can be done there. Unless I increase the volume here, the volume is a bit too low even with the volume at max in windows.

Is this something you will add or consider at a point?

I notice that a few people have requested an audio delay option as well. I think it would be neat but I've noticed that if a video or film is properly encoded it shouldn't be necessary to adjust the delay while using reclock.

[nevcairiel]

I think the main reason for me to still stick with ffdshow audio is because of the volume adjustments that can be done there. Unless I increase the volume here, the volume is a bit too low even with the volume at max in windows.

Increasing the volume in software on raw PCM is usually not a good idea, because it can easily result in clipping.
You should consider getting active speakers which can increase the volume in the analog domain.


I notice that a few people have requested an audio delay option as well. I think it would be neat but I've noticed that if a video or film is properly encoded it shouldn't be necessary to adjust the delay while using reclock.
The delay is not used because of broken encodes, its used because your hardware requires it in some cases, and nothing ReClock can do will fix it.

For me its like this:
I use HDMI out to a AV Receiver, which is then connected via HDMI to my TV.
The AV Receiver does not introduce a delay, but the TV does. The problem here is that the Audio is output without delay by the AV Receiver, but the image is delayed a bit (around 4-5 frames) by the TVs internal processing. So i set an audio delay of around 200ms, and everything is in sync again. This is quite a common problem with such setups. In the common case, the AV Receiver itself offers an option to configure the audio delay, but thats not always true.

For me, its even more complicated. The TV only introduces that delay when i play stuff at 24p, on 50 or 60Hz, there is no noticeable delay. Configuring this in the receiver is just impossible, so i use batch files to change refresh rate that also change the audio delay config in MPC-HC.

[glith]

I'm using AC3filter with about 190ms delay without reclock and every movie I watch is in perfect sync.
If I use LAV Audio instead of AC3Filter then I get very bad lip sync issues while watching 23.976 and 24hz movies.

I'd prefer using LAV audio, but I do not want to use reclock. :P

[Cudo]

It would be nice to have an adjustable crossover, I agree but some people do not use a subwoofer. They use full range front speakers to eliminate the need for it instead so IMO an option to downmix 5.1 to 5.0 would be very useful, if that's not asking too much of course. :)

[Cudo]

I'm using AC3filter with about 190ms delay without reclock and every movie I watch is in perfect sync.
If I use LAV Audio instead of AC3Filter then I get very bad lip sync issues while watching 23.976 and 24hz movies.

I'd prefer using LAV audio, but I do not want to use reclock. :P
I just did what Nevcairiel told me to, I set the audio delay in MPC-HC itself and it's been spot on ever since. Not using ReClock, just a custom resolution for 23.976 in Nvidia Control Panel.

@ Nevcairiel: I use a software tool to switch between 60 Hz, 50 Hz and 23.976 Hz that I can call from my batch files. But for some reason the Nvidia driver causes 24 Hz to be the same as 23.976 Hz until the custom resolution for 23.976 Hz is manually removed from the Nvidia Control Panel. So I have to work around this using a macro recording tool if I want to automatically switch between 23.976 Hz and 24 Hz. Do you know a better solution, by any chance?

[nlnl]

Software deinterlacing might happen for 0.37, if everything pans out.


:thanks:
And IVTC for NTSC films included?

[nevcairiel]

@ Nevcairiel: I use a software tool to switch between 60 Hz, 50 Hz and 23.976 Hz that I can call from my batch files. But for some reason the Nvidia driver causes 24 Hz to be the same as 23.976 Hz until the custom resolution for 23.976 Hz is manually removed from the Nvidia Control Panel. So I have to work around this using a macro recording tool if I want to automatically switch between 23.976 Hz and 24 Hz. Do you know a better solution, by any chance?

No, i just use 23.976 for everything. Real 24fps movies are so rare, and if i encounter one, ReClock fixes it for me.

And IVTC for NTSC films included?

No.

[Cudo]

^ Thanks again. I agree they're quite uncommon. Speeding up the audio in ReClock causes a minor loss in audio quality, though. It can easily be heard on a high end setup.

[nlnl]

Quote:
Originally Posted by nlnl View Post
And IVTC for NTSC films included?

No.

Do you have any plans for IVTC for ver. 0.xx>0.37 :)?

[nevcairiel]

^ Thanks again. I agree they're quite uncommon. Speeding up the audio in ReClock causes a minor loss in audio quality, though. It can easily be heard on a high end setup.

Its a slow-down, actually (from 24 to 23.976). Its also such a minor change that i doubt anyone can actually hear the difference. Do a double-blind test, so that your perception doesn't trick you. ;) Subjective perception is a bitch to get around. If you think it has to be worse quality, you'll most likely perceive it as lower quality as well.

I dont think its possible to work around that bug in NVIDIAs refresh rate control otherwise, both 23 and 24 map to the 23.976 mode if you create a custom resolution.

[Cudo]

Its a slow-down, actually
Yeah, sorry. lol Most people do it the opposite way.

I remember ver. 186.xx drivers didn't map the 23.976 to both 23 and 24. But their CUDA support is kind of meh.

[glith]

I just did what Nevcairiel told me to, I set the audio delay in MPC-HC itself and it's been spot on ever since. Not using ReClock, just a custom resolution for 23.976 in Nvidia Control Panel.

@ Nevcairiel: I use a software tool to switch between 60 Hz, 50 Hz and 23.976 Hz that I can call from my batch files. But for some reason the Nvidia driver causes 24 Hz to be the same as 23.976 Hz until the custom resolution for 23.976 Hz is manually removed from the Nvidia Control Panel. So I have to work around this using a macro recording tool if I want to automatically switch between 23.976 Hz and 24 Hz. Do you know a better solution, by any chance?

I'm using MediaPortal to watch my stuff.. No audio delay option in there tho. =/
(And MediaPortal has the nice built in refresh-rate changer.)

I have tried Reclock before but I couldnt really get it to work properly using HDMI. AC3filter was my saviour but cant really handle the DTS HD very well.

[6233638]

No, i just use 23.976 for everything. Real 24fps movies are so rare, and if i encounter one, ReClock fixes it for me.All films are "really" 24.0 fps.

I'm finding that an increasing number of Blu-rays are 24.0 on the disc, if that's what you meant though. (but that may be because I am in Europe)

^ Thanks again. I agree they're quite uncommon. Speeding up the audio in ReClock causes a minor loss in audio quality, though. It can easily be heard on a high end setup.If you are also oversampling the audio, I don't believe that is the case.

[nand chan]

It would be nice to have an adjustable crossover, I agree but some people do not use a subwoofer. They use full range front speakers to eliminate the need for it instead so IMO an option to downmix 5.1 to 5.0 would be very useful, if that's not asking too much of course. :)

Why don't you just use ffdshow's raw audio filter for all of this? That's what it's for. LAV Audio is just a decoder.

[Mark_A_W]

All films are "really" 24.0 fps.

I'm finding that an increasing number of Blu-rays are 24.0 on the disc, if that's what you meant though. (but that may be because I am in Europe)

If you are also oversampling the audio, I don't believe that is the case.


I've never come across a 24.0 Bluray, and I've bought/rented HEAPS, all film based (except BSG, but that's 23.976 anyway).

I agree with you about the oversampling, I don't think that the difference between 23.976 and 24.000 is audible when resampling to 96 or 192khz.

[Cudo]

^ That depends on whether it's DTS-HD MA being resampled.

[Cudo]

Why don't you just use ffdshow's raw audio filter for all of this?
Because it crashes MPC-HC alot when I'm trying to use the seekbar? :confused:

[nevcairiel]

All films are "really" 24.0 fps.

I'm finding that an increasing number of Blu-rays are 24.0 on the disc, if that's what you meant though. (but that may be because I am in Europe)


Yeah i meant coded on the disc. Its true that more discs have been 24.0 fps lately, quite an interesting development there.
I know that everything is recorded at 24.0 fps, but then the audio is mastered to match that rate exactly, so resampling it to fit a speed-up to 24.0 fps would also be lossy...

I do upsample all my audio with ReClock to 96000 Hz, so i would imagine the loss is not really noticeable. I'm still not sure if its better to watch at 23.976 or 24.000...

^ That depends on whether it's DTS-HD MA being resampled.

When its decoded, its just like any other audio. :p

[Thunderbolt8]

if everything is recorded at real 24 fps, why has the BD standard then been fixed to 23.976 fps? TV standards could have been changed as well, back then there havent been that many new flat TVs yet

[nevcairiel]

The BD standard allows both 23.976 and 24.000. 23.976 is still a relic from NTSC times, and backwards compat is apparently their main goal when developing new standards. ;)

[6233638]

The BD standard allows both 23.976 and 24.000. 23.976 is still a relic from NTSC times, and backwards compat is apparently their main goal when developing new standards. ;)I think that's probably why it seems more common on European discs, specifically of films originating in mainland Europe. (I watch a lot of French/Spanish films, though the odd British disc is 24.0 too)

Legacy in PAL regions would be 25fps/50Hz and no-one wants that (well, apart from the one or two films I have managed to get which are 1080i50) so there is no need to use 23.976 and they just go straight to 24.000.

Hopefully with the next generation of displays and video standards, we will see 24.0 worldwide. ( and possibly 48)

[dead_screem]

mp4 with track titles sample
http://www.megaupload.com/?d=KBETJMWZ

Mediainfo shows them and so does Haali. MPC-HC does as well but they only show in the audio switcher menu provided the audio switcher is enabled.

[nevcairiel]

This is stupid, as always.

Why do people use that field for the track name, instead of the dedicated atom for it? (Granted, its a quicktime atom, but shrug)
I know, probably because Haali decided to show that in the stream switcher. Formats driven by buggy implementations of it, so awesome .. not. :)

[mzso]

I think that's probably why it seems more common on European discs, specifically of films originating in mainland Europe. (I watch a lot of French/Spanish films, though the odd British disc is 24.0 too)

Legacy in PAL regions would be 25fps/50Hz and no-one wants that (well, apart from the one or two films I have managed to get which are 1080i50) so there is no need to use 23.976 and they just go straight to 24.000.

Hopefully with the next generation of displays and video standards, we will see 24.0 worldwide. ( and possibly 48)

I'm hoping for something like 120p... Displays are already capable. Studios could afford it, but so far they're too cheap. I mean friggin consumer cameras can record 60p without a problem.

[nevcairiel]

A cameras quality isn't defined by its recording speed. The High-end 4K cameras they use typically go to 48p. If there ever would be a change from 24, it would be to 48 or to 60, any higher will not happen anytime soon, if ever. Anything above 60 really isn't worth the added bandwidth/size anyway.

I would be happy enough if they get rid of interlacing and telecining, and just give us progressive video 100% of the time.

[Andy o]

I have ABXed consistently (20 consecutive times) a 1000 Hz sine and a 1001 Hz one. That means that one (me) can tell a 0.1% difference in pitch. However that's completely different than distinguishing actual content with 0.1% pitch difference, when not comparing A-B with immediate switch. I'd be skeptical that complex content (not a sine wave) would even be ABX-able with immediate switching.

24 is objectively better than 23.976, unless you like wasting 7 seconds of your life!

[Andy o]

I'm hoping for something like 120p... Displays are already capable. Studios could afford it, but so far they're too cheap. I mean friggin consumer cameras can record 60p without a problem.

I read Cameron proposing 48Hz (and not something like 60), because of easy retrofitting projectors at theaters. I don't think 120Hz would be such a gain that theaters would be willing to upgrade, even if studios did. IMO for anything above 50-60Hz diminishing returns apply.

[mzso]

A cameras quality isn't defined by its recording speed. The High-end 4K cameras they use typically go to 48p. If there ever would be a change from 24, it would be to 48 or to 60, any higher will not happen anytime soon, if ever. Anything above 60 really isn't worth the added bandwidth/size anyway.

I would be happy enough if they get rid of interlacing and telecining, and just give us progressive video 100% of the time.

Well, 120 would only double the size. :) But seriously, cinemas should provide something if they want to keep people interested. currently home cinema delivers the same if not more.

I read Cameron proposing 48Hz (and not something like 60), because of easy retrofitting projectors at theaters. I don't think 120Hz would be such a gain that theaters would be willing to upgrade, even if studios did. IMO for anything above 50-60Hz diminishing returns apply.
Well, if theaters don't upgrade Home displays can display 120 fps. They just have to shoot it at 120Hz. Cinemas will display it whatever rate they're can or want to.

[Gero]

I have a file (.m2ts) with several audio stream. Is there a way to configure the LAV splitter to make them available simultaneously? i.e. in graphedit, having a pin out for each stream? The goal is to play each stream on a different audio device.

Thanks

[Andy o]

Well, if theaters don't upgrade Home displays can display 120 fps. They just have to shoot it at 120Hz. Cinemas will display it whatever rate they're can or want to.

You think James Cameron will spend 300 million dollars for you to not pay 17 bucks per person at the theater?

[pirlouy]

120 Hz is useless for source. No need to go higher than 60 (quite eye limits).
48 Hz seems impossible, no compatibility with 60 Hz (or horrible judder)..
60 Hz ? too much data (for cd support or on-demand video)
So we had... 24Hz; sorry, won't change soon. :D

[STaRGaZeR]

120 Hz is useless for source. No need to go higher than 60 (quite eye limits).

:scared:

[TheShadowRunner]

Nev, is the On2 VP6 decoder functional?
The Lav Video Dec doesn't seem to want to accept sub-type {34564C46-0000-0010-8000-00AA00389B71} 4CC:FLV4, even though On2 VP6 should be able to handle them I believe. (sample (http://www.square-enix.co.jp/music/sem/page/xenogears_orchestra/swf/soundsbgm/pv.flv))

[clsid]

That sample plays fine here with LAV Video.

[TheShadowRunner]

That sample plays fine here with LAV Video.

I guess you're using LAV Splitter which I suspect is presenting another sub-type to LAV Video Dec. I use FLVSplitter.
Could you tell me the sub-type on the LAV Video filter input pin on your working graph?

[JarrettH]

Does the apply DRC effect work in lav audio? If so, it seems fairly mild at 100% and yes I am using it on an AC3 file. :thanks:

[nevcairiel]

I have a file (.m2ts) with several audio stream. Is there a way to configure the LAV splitter to make them available simultaneously? i.e. in graphedit, having a pin out for each stream?

Thats not possible at this time.

Nev, is the On2 VP6 decoder functional?
The Lav Video Dec doesn't seem to want to accept sub-type {34564C46-0000-0010-8000-00AA00389B71} 4CC:FLV4, even though On2 VP6 should be able to handle them I believe. (sample (http://www.square-enix.co.jp/music/sem/page/xenogears_orchestra/swf/soundsbgm/pv.flv))
It only accepts a subtype of VP6F (VP6 Flash). I can probably make it accept FLV4 as a subtype, if thats documented somewhere to be common usage.
Some quick googling reveals that it should map to the same, so i'll try that later.

Does the apply DRC effect work in lav audio? If so, it seems fairly mild at 100% and yes I am using it on an AC3 file. :thanks:
I have no idea how strong it really is, i just expose the setting of the AC3 decoder in ffmpeg.

[mzso]

120 Hz is useless for source. No need to go higher than 60 (quite eye limits).
48 Hz seems impossible, no compatibility with 60 Hz (or horrible judder)..
60 Hz ? too much data (for cd support or on-demand video)
So we had... 24Hz; sorry, won't change soon. :D

I guess you never used a 60 hz crt.
You think James Cameron will spend 300 million dollars for you to not pay 17 bucks per person at the theater?
Oh he would get his 17 bucks. You just wouldn't get full framerate everywhere. :)

[nevcairiel]

I guess you never used a 60 hz crt.

60Hz CRT are so bad because its CRT, not because the source is only 60 fps. Its a technical problem, it has nothing to do with the source refresh rate.


Oh he would get his 17 bucks. You just wouldn't get full framerate everywhere. :)

Why switch to something that wouldn't be used anywhere, except in the homes of people that spent a whole lot of money on high-end gear? :p

Home cinema equipment mostly already supports 60 fps, so if anything, that would be something to go to. If they want to go beyond what we can do now, i rather see increased resolutions, and a focus on higher bitdepth panels.
And there is of course the topic of chroma subsampling. The next disc/streaming format should rather focus on improved resolution and at least 4:2:2 chroma, instead of something totally overkill like 120 fps. :p

Luckily, the film makers are not as insane, and will never go to 120 fps. You can argue all you want, its just not going to happen. Going 48 would be a good idea, sadly its not compatible with existing techs, as 48 is such a odd and unsupported resolution.

Anyhow, this discussion is totally off-topic for this thread, and if you want to continue, please take it elsewhere. :)

[sneaker_ger]

The problem with higher frame rates is also an artistic question. The "look and feel" of the movies that we are used to, could change in a way that audiences might not accept. Anyone see "Public Enemies"? I couldn't get used to its look. Or the higher framerate could be limited to scenes that need it, like pans. Higher resolutions, better subsampling (or rather the lack thereof), higher bitdepth are more likely to be accepted (and the first and the latter could be marketed the best, I presume.)

[mzso]

60Hz CRT are so bad because its CRT, not because the source is only 60 fps. Its a technical problem, it has nothing to do with the source refresh rate.
Its bad because its crt and 60hz. With 100 hz crt's people didn't have problems, because that's close to the limit of perception.


Why switch to something that wouldn't be used anywhere, except in the homes of people that spent a whole lot of money on high-end gear? :p

Home cinema equipment mostly already supports 60 fps, so if anything, that would be something to go to. If they want to go beyond what we can do now, i rather see increased resolutions, and a focus on higher bitdepth panels.
And there is of course the topic of chroma subsampling. The next disc/streaming format should rather focus on improved resolution and at least 4:2:2 chroma, instead of something totally overkill like 120 fps. :p

Well 120hz and such seems quite common to me with the 3d push and all. Also it very quickly can become the norm if manufacturers fire up their marketing machines. People keep buying new tv even when there are 99% the same as the previous generation....

I just don't see why not make the leap and be done with it for good.

4:2:0 is just stupid standard isn't it? They could just stop butchering the colors. Hell it might work well on 50gb blu ray size with a good encoder The source is full color I suppose.

Speaking of colors I wonder what resolution would be enough. 30 bit? 36?

[nand chan]

Why is everybody so hung up about the framerates?

The REAL problem with cinema movies and the Blu-ray process is the current limitation to the BT.709 primaries. So many displays can display significantly larger gamuts, with higher end projectors reaching up to 180% NTSC (movies are at 72%)! Why should we have to go to theaters to get the unbutchered masters?

The only problem is the Blu-ray process. If we started switching to xvYCC or ACES or ProPhoto RGB or whatever, the whole process would be much, much simpler. That, and of course, 16-bit video to eliminate the banding this would introduce.

And, of course, stupid chroma sub-sampling. If they cut out all of the completely useless DTS, AC3, MPEG-2 and whatever compatibility tracks and just switched to 16-bit 4:4:4 H.264 with ACES primaries + APE / TTA (or possibly FLAC), we would get pretty much the same quality while making full usage of the available 50 GB of a dual-layer Blu-ray.

Or maybe even glass silica discs in the near future, with which we could basically store RGB32 with no problems.

Oh, and we could fix the framerate problem by allowing variable frame-rates, eg. speeding up to 120 Hz for pans but slowing down to 24 Hz or whatever for scenes with a lot of static imagery or motion blur.

[nevcairiel]

Anything, but not VFR, unless you build a screen that supports VFR as well.

[SamuriHL]

Anything, but not VFR, unless you build a screen that supports VFR as well.

You got that right! My screen takes on average about 2-3 seconds to switch. Forget that!!

[nevcairiel]

LAV Filters 0.36

LAV Splitter
- Improved support for ADTS AAC
- Fixed MPEG-2 timestamps in VOB
- Fixed MPEG-2 frame rate reporting for telecined/interlaced

LAV Video
- Integrated NVIDIA CUVID Hardware decoder
- Performance enhancements in 10bit decoding
- Support for Flash Screen Video (FSV1)
- Support for Apple ProRes
- Improved MPEG-2 playback
- Improved A/V sync with VC-1 and MPEG4-ASP


Download: Installer (both x86/x64) (http://files.1f0.de/lavf/LAVFilters-0.36.exe) -- Zips: 32-bit (http://files.1f0.de/lavf/LAVFilters-0.36.zip) & 64-bit (http://files.1f0.de/lavf/LAVFilters-0.36-x64.zip)

It was time to get this out for good. You can only do so much testing yourself, at some point you just gotta get it out there. ;)

The biggest change in this release is obviously the integration of the CUVID decoder. To embed the decoder, LAV Video was basically rewritten to 80-90%, because it was initially not designed to support more then one decoder properly.
A big change like that always has the potential to break stuff, but also to improve things, so we'll see how it goes. :)

The CUVID decoder supports full software fallback for unsupported H264 profiles (the hardware only supports High Profile, which means 4:2:0 8bit), as well as 4:2:2 MPEG-2. You can however not use another software decoder when CUVID fails, but seriously who would want that? :D
In addition to porting the existing CUVID functionality, there is a new option called "Force Deinterlacing", which will force the deinterlacer to process all frames, even if marked progressive. This can be used to deinterlace videos that were encoded as progressive but actually are weaved interlaced fields.

Note that LAV CUVID will basically be discontinued now, all its functionality is available through LAV Video. Only missing is MPEG4-ASP decoding through CUVID, but i don't believe the extra effort is really worth it.

Anyway, have fun with this release, and report bugs and regressions!

[SamuriHL]

Awesome. Thanks, Nev! Now let's see if we can get Matt to update MC16 to use this at least in place of the existing CUVID.

[ney2x]

I am at work right now and I'm only using a laptop so it's impossible for me to test the latest release. My questions are, does LAV CUVID ends with the version 0.13? Does Arcsoft dts*.dll workaround still recommended? Thanks.

[Xaurus]

Thank you nevcairiel! I've said goodbye to LAV CUVID now.

Q: When does the multi-threaded function work? I am trying a h.264 clip with hardware disabled and threads set to 6... but in task manager it only use 1 thread.

Q: There is no way yet to determine whether hardware or software decoding is in use?

[nand chan]

- Performance enhancements in 10bit decoding

Great to hear. How much of an increase have you been measuring, on average?