LAV Filters - DirectShow Media Splitter and Decoders

[egur]

I use LAV - it's all around more robust, handles many more containers, fixes some streams and more.

Haali supports siff files (3D H264) in a way that (only) CoreMVC expects and has much better seeking in TS files.
Haali has serious issues with TS files that LAV doesn't - bad time stamps and bad aspect ratio.

So neither is perfect (yet :) ) but LAV is much better in my opinion.

Kudos to Nev on both his effort and result!

[FDisk80]

I use LAV - it's all around more robust, handles many more containers, fixes some streams and more.

Haali supports siff files (3D H264) in a way that (only) CoreMVC expects and has much better seeking in TS files.
Haali has serious issues with TS files that LAV doesn't - bad time stamps and bad aspect ratio.

So neither is perfect (yet :) ) but LAV is much better in my opinion.

Kudos to Nev on both his effort and result!

Nev תותח עולם! :D

[rlevis]

Can no one help with my query regarding AC3 audio being much quieter than AAC+ with LAV and not other audio decoders? Is it perhaps something to do with AC3 not being mixed down to 2 channels properly?

Hi.

I just switched to LAV Audio Decoder for MediaPortal, mainly due to the audio delay feature which I need for my new TV. I was using Monogram for AAC+ and ffdshow Audio Decoder for everything else, but Monogram doesn't have an audio delay feature.

Unfortunately the AC3 audio is much quieter now, but AAC+ is normal volume, and this makes it a hassle when changing channels. In NZ we have some digital channels with AC3 surround and some AAC+ stereo. We only have stereo speakers.

I tried enabling the DRC audio compression checkbox but it hasn't made any difference I can notice.

I don't suppose there is any method to resolve this?

[nevcairiel]

Can no one help with my query regarding AC3 audio being much quieter than AAC+ with LAV and not other audio decoders? Is it perhaps something to do with AC3 not being mixed down to 2 channels properly?

LAV does no mixing at all. If you require mixing, you'll either have to wait a version or two until it gets implemented, or combine it with ffdshow to use its mixer.

Not doing proper mixing can certainly cause volume differences.

[pururin]

LAV splitter is better. Haali is old and not get updated for a long time.

However, Haali splitter has really nice tray icon that shows streams bit rate
and support ordered chapter which is important to some viewers.

These are the benefits over LAV splitter. (for now?)

[BetA13]

hy there....

@ nevcairiel

I just discovered that i cant watch my recorded game videos with LAV...I think it worked before...
I there a chance that u can add an codec to ur LAVFilters?
Im using the RTV1 Codec...its a very good one...(riva tuner video i think) its implented in Afterburner...

Here is the Codec if u need it:

http://www.mediafire.com/?l9gw2i7cymdy611


Thanks a lot and keep up the GOOD work :)

schöne grüße ;)

[_gl]

LAV splitter is better. Haali is old and not get updated for a long time.

For me the exciting thing about LAV is that it tries to unify everything into a few codecs. I'm really tired of 'codec hell' (this splitter doesn't work this this codec but the other one doesn't create Explorer thumbnails, blah blah). Especially if you author video as well as watch, it's a constant configuration battle.

So even if LAV doesn't do a few things yet, let's support Nev as much as we can to gradually get them in there.

[cengizhan]

LAV splitter is better. Haali is old and not get updated for a long time.

However, Haali splitter has really nice tray icon that shows streams bit rate
and support ordered chapter which is important to some viewers.

These are the benefits over LAV splitter. (for now?)

haali is still updated by coreavc. also it is still the fastest mkv splitter.

[kolak]

There is a bug in DNxHD 10bit decoding- on high contrast areas it creates artefacts (outline- looks similar to mosquito noise). It was reported to ffmbc developer and fixed, but looks like still present in ffmpeg.

[wanezhiling]

http://www.mediafire.com/?iuinywzgt2geh5c
nev, lav splitter 0.50.2-d6cbba7 cannot recognize second stream.

[nevcairiel]

http://www.mediafire.com/?iuinywzgt2geh5c
nev, lav splitter 0.50.2-d6cbba7 cannot recognize second stream.

Actually, it can, it just doesn't show them because they're in a different program.

Speaking about programs, i was meaning to improve support for that at some time in the future, but i'm not sure about the best approach.

The easiest solution would be to just show all streams, and add a PID identifier to the stream selector (like MPC-HC Mpeg Splitter)
Additionally, i could add a 4th "stream" category which lists the programs. Selecting one of those programs would then switch all 3 stream types to that program. This allows full manual stream control, as well as an automated way to switch all streams at once.

From the implementation stand-point, this is also the easiest solution.
Thoughts?

[wanezhiling]

Actually, it can, it just doesn't show them because they're in a different program.

:thanks: for quick reply.
Additionally, i could add a 4th "stream" category which lists the programs. Selecting one of those programs would then switch all 3 stream types to that program. This allows full manual stream control, as well as an automated way to switch all streams at once.
:) I prefer this.
Of course, its up to you

[Mercury_22]

Actually, it can, it just doesn't show them because they're in a different program.

Speaking about programs, i was meaning to improve support for that at some time in the future, but i'm not sure about the best approach.

The easiest solution would be to just show all streams, and add a PID identifier to the stream selector (like MPC-HC Mpeg Splitter)
Additionally, i could add a 4th "stream" category which lists the programs. Selecting one of those programs would then switch all 3 stream types to that program. This allows full manual stream control, as well as an automated way to switch all streams at once.

From the implementation stand-point, this is also the easiest solution.
Thoughts?

I too vote for the last one (4th "stream" category) also can you please add the installer to the build.bat I'm already using my modify bat for that but it will be nicer to have an "official" one :D

[nevcairiel]

I too vote for the last one (4th "stream" category)

There is no "last one", its one total idea.
Show all streams of all programs, and in addition show "virtual" streams that switch audio/video at the same time.

[Mercury_22]

There is no "last one", its one total idea.
Show all streams of all programs, and in addition show "virtual" streams that switch audio/video at the same time.

Yes for the additional "virtual" streams cause let's faced when switching programs in 99.99% of the time we need the both new video and audio streams

[jmone]

LAV does no mixing at all. If you require mixing, you'll either have to wait a version or two until it gets implemented, or combine it with ffdshow to use its mixer.

[flame suit on]...or do the mixing in the player, where I think it belongs! [/flame suit off]

[pankov]

There is no "last one", its one total idea.
Show all streams of all programs, and in addition show "virtual" streams that switch audio/video at the same time.

Nev,
what's the point in showing all streams from all programs at the same time? Personally I think it will be better to just show a list of all programs and only the streams that belong to the currently selected one.
Does anybody need picture from one program and sound from another? When?

[nevcairiel]

Nev,
what's the point in showing all streams from all programs at the same time? Personally I think it will be better to just show a list of all programs and only the streams that belong to the currently selected one.
Does anybody need picture from one program and sound from another? When?

Doing that adds quite some complex situations that would need to be handled, which would mean i would probably never implement it because i don't care about that feature. :P
Not to mention that players may extract the available streams at start and never update their list later on, because there is no situation right now where it would ever change.

Hiding information from the user is usually not a good way to go, maybe a file is just broken and the streams of all programs are meant to go together, it just adds flexibility to the whole thing. Granted if you have a file with several programs, the list may get long, but i really haven't seen files with more then 2-3 programs in them.

[pankov]

I see - it's more like a technical issue.
I just thought it would be better to mimic the way that TVs and STBs work - they offer you only the streams that belong to the currently selected program despite the fact that there are more programs/channels on the same transponder ... up to 7-8 different ones)
I think multi-program/channel files are produced by recording the full transponder stream. This will lead to 6-8 channels in the same file (at least that's the average number of SD channels we have here). Multiply this by 2-3 and the result will be something ~15 ... which is not that bad.
I think I can live with the solution you offer.

[egur]

I did some testing on IvyBridge today.
System:
* CPU: i7 class engineering sample (E0) @ 2.6GHz.
* DDR3@1333 sodimm.
* Window 7
* 2712 graphics driver
* ZoomPlayer
* CoreTemp
* LAV splitter
* Decoders: LAV 0.50.2 and my own latest ffdshow build.

Good news:
* Managed CrowdedRun (4K@50fps, ~122mbps) with 30% CPU@3.2GHz using both ffdshow and LAV. libavcodec (ffdshow) took 85% CPU@2600 (jumps to 2700).
* This clip is the worst case scenario - very high bitrate, high resolution, high frame rate - stresses all the subsytems (memory, decoder, CPU). Most 4K clips have 1/4 of the bitrate and half the frame rate.
* Other clips played fine no surprises so far.
* A transposed 720p clip (720x1280) played very well. SandyBridge's QS can't play it since the line count is >1080.

The bad news:
* The reference board used didn't have proper cooling and the CPU hit 103C (SW or QS). At these temperatures it activates throttling to cool itself down. This might explain the high CPU usage. I need to rig it with a better fan and test again.

Even with these far-from-optimal conditions the video playback was smooth.

Also, started working on adding HW video processing. Some stuff already works but some issues are too severe for a proper release :(
I'll commit to SVN my changes soon so Nev can start playing with it.
Added:
* Deinterlacing (half/full rate output)
* Detail filter
* Denoise filter

Not working:
* 50i sources
* Telecined sources

Didn't add procamp (Hue, Saturation, Brightness, Contrast) yet, I'll add it too (very simple to do but not so useful).

Update
Fixed the cooling solution and QS behaves the same. libavcodec now raises CPU frequency to 3.2GHz and uses all cores at 82%.

[SamuriHL]

Very nice! My i7 is mostly built but I found that my mobo is defective with 2 dead dimm slots so until the replacement gets here tomorrow I can only use 8 gigs of the 2133 I got. I haven't yet tried the igpu but I want to see how good virtu is at some point.

Sent from my Xoom using Tapatalk 2

[Sebastiii]

I did some testing on IvyBridge today.
Update
Fixed the cooling solution and QS behaves the same. libavcodec now raises CPU frequency to 3.2GHz and uses all cores at 82%.

Hi :)

Great news and how did you fix the cooling solution.
I have hear that new i7 Ivy Bridge is higher (20°C) than sandy bridge one.

Thanks.

[SamuriHL]

When overclocking, yes, that's true. I had heat issues with mine for the first hour I had it running. Turns out one corner of the HSF wasn't fully seated. I'm NOT looking forward to having to move everything to the replacement mobo for that very reason.

[kkrull]

Is anyone else working on native BluRay WMC playback?

Hi,

The playback of main titles from BDMV folders using the index.bdmv file is working perfectly for me in WMP/WMC (Win7) - many thanks for this feature.

The only issue I can't figure out is how to get Windows 7 Media Center to recognise the index.bdmv file so that it can start playback from the Movie Library.

If I double click on an index.bdmv file from explorer it will start fine in WMC. Usually WMC will recognise a BDMV folder and throw an error when I try to play saying I need a 3rd party blu-ray player. Changing the name of the BDMV folder doesn't work; doing this together with registering the .bdmv extension as a video file type does allow WMC to 'see' the index.bdmv file but playback fails.

Any ideas? Getting ripped blu-ray folder playback working to avoid the need for remuxing to mkv and without the need for Arcsoft or PowerDVD would be great!

Cheers

[SamuriHL]

What exactly does that have to do with LAV Filters?

[Tacio]

Also, started working on adding HW video processing
Will it only concern IvyBridge? Or SandyBridge as well?

[egur]

Will it only concern IvyBridge? Or SandyBridge as well?

Both supported. Future CPUs will also be supported.

[SamuriHL]

Sweet. That'll be nice.

[jakmal]

Both supported. Future CPUs will also be supported.

Will it be self-contained, i.e, override the driver settings as put in the Intel Graphics Control Panel ?

[egur]

Will it be self-contained, i.e, override the driver settings as put in the Intel Graphics Control Panel ?

I've verified it empirically, it ignores the control panel settings (as it should).

[6233638]

It doesn't seem that anyone has reported this, but 0.50.2 breaks DVD playback with madVR.

I have madVR set as the video renderer, and ReClock set as the audio renderer in MPC-HC.

External filter list is configured as: LAV Splitter Source LAV Audio Decoder ffdshow Audio Processor (to downmix audio) LAV Video Decoder ffdshow Video Decoder (fallback for DVD playback, which LAV does not yet support)

[nevcairiel]

The only thing from that list even active for DVDs would be LAV Audio, which has only gotten very minor internal changes in the last version.
I never understood when DVD with madVR worked and when it breaks, it seems completely arbitrary to me, so i don't consider this a bug, because it probably isn't caused by any of the changes but some other random thing.

[Sebastiii]

When overclocking, yes, that's true. I had heat issues with mine for the first hour I had it running. Turns out one corner of the HSF wasn't fully seated. I'm NOT looking forward to having to move everything to the replacement mobo for that very reason.

Ok thanks but for now i hesitate to get new cpu lol

[andyvt]

I've verified it empirically, it ignores the control panel settings (as it should).

Does it depend on what's selected ("Application" v. "Driver" Settings)?

[Plutotype]

I'll let you know in two weeks. :p
SNB can already mostly do it, should be just fine with IVB.

Hi Nev,
Did you manage please to test the Ivy Bridge platform with madVR and high bitrate 1080p material? Is discrete graphics not needed anymore for a HTPC?

Thanks
Pluto

[RBG]

egur

And what about hw scaler, do you intend to implement it too?

[SamuriHL]

Ok thanks but for now i hesitate to get new cpu lol

I have no regrets. The issues I'm facing will get fixed, and when they are, this machine is going to be INSANE. In a month or two I'll add a 680 and I'll have a machine that'll last me a good number of years.

[egur]

Does it depend on what's selected ("Application" v. "Driver" Settings)?
It ignores the "driver custom setting". This is the behavior of the Intel Media SDK (which is a good thing) since VPP features are also used for transcoding. The directshow filter (e.g. LAV video decoder) or by some means the player should control these values and expose a GUI to the user.

egur
And what about hw scaler, do you intend to implement it too?
One thing at a time, the deinterlacing feature is far from completion.
A new reason to add HW scaling is for downscaling 4K video when the screen is smaller it will reduce copying significantly. I believe this should also be controlled by the player and/or filter.

[egur]

Committed my changes to SVN (r50).
This version is for testing/development purposes only and should not be used in production SW.

[xv]

Found a little problem: LAVSplitter cannot handle sample rate changes in DV (AVI Type-1 DV) files: For example some recordings start with 32khz 12bit for the first few frames because it was some previous recording on the tape, then the audio is actually 48khz/16bit. Microsoft DV Splitter has no problem with that, but LAVSplitter won´t update the format and mess up audio. Could this please be fixed?

[jakmal]

Hendrik,

When using LAV Video Decoder (in CUVID mode) + madVR (with madVR set to do deinterlacing), is there any advantage to selecting 'High Quality Processing' in the LAV Video Decoder configuration?

[nevcairiel]

Hendrik,

When using LAV Video Decoder (in CUVID mode) + madVR (with madVR set to do deinterlacing), is there any advantage to selecting 'High Quality Processing' in the LAV Video Decoder configuration?

There is no advantage in turning it off.

Anyway, it only affects deinterlacing, if you have LAV set to "Weave", it won't change anything because then the renderer does the job.

[wanezhiling]

:)
nev, what's the difference between LAV's CUDA deinterlacing and madVR's DXVA2 deinterlacing?

Could madVRs beat yours if I disable HQP? :D (just kidding)

[nevcairiel]

nev, what's the difference between LAV's CUDA deinterlacing and madVR's DXVA2 deinterlacing?

None, its the same. Both ask the hardware to do deinterlacing


Could madVRs beat yours if I disable HQP? :D (just kidding)
If you disable HQP, madVRs or EVRs will be better. But why would you turn it off?
I should just remove that option.

[wanezhiling]

None, its the same. Both ask the hardware to do deinterlacing
Thanks.

http://forum.doom9.org/showpost.php?p=1536640&postcount=10603
(2) Using DXVA2 deinterlacing means that potentially all the nasty GPU algorithms like noise reduction and edge enhancement are introduced to madVR.
Same is LAVs?

I should just remove that option.
Good idea.:p

[nevcairiel]

Same is LAVs?

Yes, although CUVID only works with NVIDIA, and NVIDIA has all these settings off by default, so its your own fault for turning them on. :)

[jakmal]

So, I have some interesting results in this area with the GT 540M (Maybe others can try to reproduce with other GPUs?):

Configuration 1 :

LAV CUVID + Adaptive HQP Deinterlacing + Aggressive Deinterlacing
madVR : Force deinterlacing if in doubt

Configuration 2 :

LAV CUVID + Weave (None) / HQP Turned Off + Aggressive Deinterlacing Ticked (should have no effect?)
madVR : Force deinterlacing if in doubt

For renethx's La Travita 1080i60 H.264 clip (http://www.4shared.com/video/1McC98V-/La_Traviata__2010__2_min_HD_p_.html) :

GPU usage in configuration 1 goes more than 95% and plenty of dropped frames throughout. in configuration 2, stays around 85 - 92%, no dropped frames -- the render and presentation queues remain almost full all throughout.

[nevcairiel]

Its not really interesting, rather expected actually.

The first configuration will need more power because it has to download 60 full frames from the GPU, and then upload them back in madVR.
The second configuration only downloads 30 frames from the GPU (and uploads those back in madVR), and the deinterlaced 60 progressive frames never leave the GPU again (its happening all inside madVR)

This can cause quite a bit of extra load on memory constraint systems, like mobile GPUs. Even the second case is already running very close to capacity.

[wanezhiling]

Its not really interesting, rather expected actually.

The first configuration will need more power because it has to download 60 full frames from the GPU, and then upload them back in madVR.
The second configuration only downloads 30 frames from the GPU (and uploads those back in madVR), and the deinterlaced 60 progressive frames never leave the GPU again (its happening all inside madVR)

This can cause quite a bit of extra load on memory constraint systems, like mobile GPUs. Even the second case is already running very close to capacity.
:thanks:Useful info to me.


nev I have a question:
Pure DXVA costs less GPU performance compared to those cb/qs etc. because it needn't do a copy-back step, but how to explain the same result about cpu usage(naitve < cb/qs)?

[nevcairiel]

nev I have a question:
Pure DXVA costs less GPU performance compared to those cb/qs etc. because it needn't do a copy-back step, but how to explain the same result about cpu usage(naitve < cb/qs)?

When you copy from GPU to CPU, and then back from CPU to GPU, obviously the CPU also has to do some work.