LAV Filters - DirectShow Media Splitter and Decoders [Archive] - Page 79

nevcairiel

19th June 2011, 19:07

It can. As long as 'VC-1 timestamp correction' is disabled. But then m2ts playback is choppy which requires the correction. If 'VC-1 timestamp correction' is enabled, m2ts playback is smooth, but MKV is choppy. Seems as if frames are being swapped in both "error" cases or something like that.

I'll add ffdshow DXVA to the list of broken codecs then, which will cause the "auto" setting to enable it for m2ts and disable for MKV.

Andy o

19th June 2011, 20:11

@nevcairiel

Just a quick question. Is there a plan for AC3 (SPDIF Encode mode) like the ffdshow? A friend of mine ask me...

You can do that with ReClock, or (I think) with ffdshow as audio processor, both of which will still let you use LAV as a decoder.

zipi

19th June 2011, 23:18

Is there any reason why LAV Audio Decoder outputs 32 bit float on MP3s (or at all for that matter when connected to a real hardware audio renderer) ?

how can one disable floating point output ?

I don't think you can, but what do you mean by "real hardware audio renderer"? Example?

any real soundcard audio renderer, not an intermediate filter/processor.
With many pro cards,if you feed them 32bit float, you'll get a distortion storm guaranteed to kill your gear & hearing.

nevcairiel?

nevcairiel

19th June 2011, 23:23

Your so called "pro cards" should get a driver then that does either not accept float input, or just converts it to integer before sending it over the DACs.

Forcing a specific output format may be possible one day, but since it degrades quality, its not on the top of the list.

zipi

19th June 2011, 23:30

I cannot say I am too worried (or understand your logic tbh) about "losing quality", on mp3 decoding, by not using 32 bit float output for mp3s.
Not to mention I have not seen LAV Audio decoder do it with Wavpack for example, using the same cards.

btw about the "so called" part of your reply :):
I actually stumbled upon this problem by accident setting up a system with LAV installed,
As I build audio oriented systems & dealing with a lot of professional cards, I thought it is worth reporting it to you, so whether it gets fixed or not, I have zero interest in it either way ...

Thanks for replying, all the best.

Monamona

19th June 2011, 23:52

This is all in the hands of ffmpeg. There have however been some fixes coming in since the last time i updated ffmpeg, and i'll be merging those fixes for the next version.

Do you have a sample file which shows the problem, so i can see if things improved?

I uploaded a sample as follows.

https://rapidshare.com/files/2471309460/sample.mp4
(200MB, sorry...)

I put chapters at the beginning of scenes (like opening main-part...) so that you can see gaps of 2 frames.

Portioli

20th June 2011, 00:44

sorry if this is offtopic but i would like to make a question about this hd audio discussion.

as long as lav audio can decode Dolby TrueHD (and DTS-HD MA with the TMT5 trick)
is it possible to send hd audio uncompressed via hdmi ( multi-channel PCM ) to old AV Receivers that do not support HBR decoding (dts-hd ma + DD TrueHD) ?

are the audio drivers/card drivers or the hdmi cable of a pc, capable to send 6 or 8 channels of losseless audio to the receiver?

Are analog outputs the only solution?

thanks in advance

Andy o

20th June 2011, 01:05

any real soundcard audio renderer, not an intermediate filter/processor.
With many pro cards,if you feed them 32bit float, you'll get a distortion storm guaranteed to kill your gear & hearing.

nevcairiel?

Why must you do it this way? It can't be latency (for movies/music it shouldn't be an issue). Isn't it easier to just use a proper renderer to output 24-bit or whatever to your card?

It's what LAV audio would do, btw, if you could "force" it to output integer from lossy sources (correct me if I'm wrong, Nev).

Andy o

20th June 2011, 01:21

as long as lav audio can decode Dolby TrueHD (and DTS-HD MA with the TMT5 trick)
is it possible to send hd audio uncompressed via hdmi ( multi-channel PCM ) to old AV Receivers that do not support HBR decoding (dts-hd ma + DD TrueHD) ?
Sure, as long as you have a full HDMI audio device. ATI/AMD 4000 and newer, Intel G35, G45 and Arrandale/Clarkdale and newer (with proper mobo support), Nvidia 460GTX and lower 400 series, and their 500 series counterparts. Some older Nvidia integrated chipsets too, like the 9300, 9400, 8200, 8300.

Portioli

20th June 2011, 08:41

Thanks for this Andy

Blight

20th June 2011, 11:35

Hi,
I have an MOV file which appears to have PCM (uncompressed) audio (download here : http://games.on.net/file/30341/Borderlands_-_Australian_launch_trailer).
When using the 'LAV Source Splitter' filter, it exposes two sub-types:
{34616D69-0000-0010-8000-00AA00389B71} // QDesign Music 2
{AFBC2343-3DCB-4047-9655-E1E62A61B1C5} // E-AC3

Since neither is the actual format, any audio decoder other than the 'LAV Audio Decoder' fails to connect.

nevcairiel

20th June 2011, 14:12

The first type is called "IMA4". Its an ADPCM codec, and ffdshow should support it.
The second type is not E-AC3, its a LAV Splitter internal type. Someone mislabeld it in some ZP profile.

mark0077

20th June 2011, 22:13

nev, sorry if it was mentioned in an earlier thread but should I be able to seek in mkv's with the current lav splitter. I have alot of mkvs I created from DVDs using MakeMkv in the past.

I can seek using the internal mpc-hc or haali mkv splitter but with lav splitter, when I seek to a certain location, the splitter jumps to the previous / next chapter, rather than the specific times I want to seek to.

nevcairiel

20th June 2011, 23:09

I can seek just fine in any MKVs i have.
Try turning off "Fast seek(on key-frame)" in the MPC-HC "Tweaks" options, maybe it helps your case, its sometimes somewhat buggy (MPC-HCs fault)

JarrettH

20th June 2011, 23:30

Can you tell me what is more correct for my setup?
http://www.emu.com/products/product.asp?product=10447

It seems LAV Audio outputs most audio streams in 32-bit, so I'll begin with that...

LAV Audio Output 32-bit --> PCM 32-bit --> AC3 Filter to downmix to 2.0 and using PCM 24-bit Output (this matches the EMU 0404 24-bit output rate)

OR

LAV Audio Output 32-bit --> PCM 32-bit --> AC3 Filter to downmix to 2.0 and using PCM 32-bit Float Output (EMU 0404 converts this to 24-bit output)

I think the first scenario makes more sense, do you? :thanks:

Mark_A_W

21st June 2011, 06:24

PeQuE

21st June 2011, 07:56

Hi nev...

I know you are on vacation but... Could you think about the possibility of adding audio delay feature to Lav Audio Decoder? I only miss this to finally get rid of ffdshow... :(

Thanks a lot for considering...

mindbomb

21st June 2011, 09:03

CiNcH

21st June 2011, 16:33

I'll add ffdshow DXVA to the list of broken codecs then, which will cause the "auto" setting to enable it for m2ts and disable for MKV.

OK.

Just a small addition. Haali + ffdshow DXVA is also fine for both, VC-1 within m2ts and MKV. I don't believe that it checks the downstream filter.

nevcairiel

21st June 2011, 16:57

Maybe it always does it for m2ts and never for MKV, only way to confirm is to run it through some other (commercial) decoders which require the opposite settings in LAV Splitter.

JarrettH

21st June 2011, 17:08

heh, i went through this type of thing before myself.

Firstly, you should use ffdshow instead of ac3 filter, iirc, ac3 filter freaks out when you play things with more than 6 channels.

second, it is better to downmix in 32 float.

Thanks, actually I have ffdshow, but I always preferred the downmixing in AC3 filter. I'll give ffdshow another try though, I'd love to get rid of that additional application!

Is the Processing tab in ffdshow the input?

Dogway

21st June 2011, 17:22

FYI after 2 months I discovered how to demux the file I was having problems with in this post (http://forum.doom9.org/showthread.php?p=1491561#post1491561), avc+aac in a QT muxed ".mov". Mov extension doesn't show up in lavsplitter input formats, so simply rename to mp4 and everything will work as expected. Its a nice workaround until nev gives support :cool:
lavsplitter through graphstudio could automatically detect it as mp4, but through mpc the video screen was just messed freeze so I guess it detected as mov

mark0077

21st June 2011, 19:16

I can seek just fine in any MKVs i have.
Try turning off "Fast seek(on key-frame)" in the MPC-HC "Tweaks" options, maybe it helps your case, its sometimes somewhat buggy (MPC-HCs fault)

Thanks nev, that seemed to help yes. However seeking large distances in my mkv files with lav splitter has some issues.

If I seek for example 1 hour forward in an mkv, the image just seems to stop / freeze, and the audio continues in the background. Sometimes if I seek say about 10 minutes forward, I see like about 1000 frames flash past over the space of a second or two, then I'm at my desired location.

Does the splitter actually still send all of the frames from my start to finish location to the decoder, it seems that way at present. I don't get instantly moved to the new location like with haali for example. Am I still hitting more mpc-hc problems or could it be lav splitter?

Try opening a largish, 2hr mkv and seek to almost the end to see if you can reproduce, if you'd like? Definitely something not right, let me know if you need any more info if you're interested in trying to reproduce etc.

EDIT: Actually when I disable the lav cuvid decoder, I can't reproduce any of the issues above, I can seek almost instantly in these files. Maybe its lav cuvid thats causing the issue? Let me know if I should repost over there instead... I have lavcuvid set to decode h264, vc-1, mpeg2, deinterlacing set to None, 50p/60p, Field Order: Auto, Settings are "Use stream aspect ratio" on and "Use HQ DXVA processing" off, with Output Format: YV12

Using GTX 295, Windows 7 64bit, mpc-hc 32bit with lavcuvid and lavsplitter/lav audio.

nevcairiel

21st June 2011, 19:50

It doesn't seen all frames, of course, but seeking is sometimes a bit inaccurate so that it ends up at a few seconds earlier (usually never later), and those frames are delivered to the downstream filters completely, causing those flashing frames in LAV CUVID, because it doesn't discard them like other decoders. I should fix that in LAV CUVID, as well as improve seeking accuracy in some file formats.

On your original issue however, your MKVs seem weird. Usually a MKV should have a Cue Point every few seconds, those cue points are used for the "fast seeking" in MPC-HC, if MakeMKV only puts them on chapters, i consider that a bug. Sadly MPC-HC is rather stupid, and it always seeks to cue points when that option is one, even if its time is many seconds or even minutes away.

mark0077

21st June 2011, 19:51

ah okie, no problemoz

RealSnoopyDog

22nd June 2011, 10:24

Hi nevcairiel, i started compiling your splitter and got to the point where i need to compile ffmpeg. This not so easy (compiling ffdshow is more easy) but anyway. I use MinGW and got it working, the only thing is: what parameters should i use for the "configure" script of ffmpeg. I found somewhere that i should use the options "--enable-memalign-hack" if the destination is Windows. So i use "--enable-memalign-hack --enable-shared --cpu=x86". Is this what you also use? You know, MinGW runs sooooo sloooowly..... I get grey hair when i trial-and-error this ;)

nevcairiel

22nd June 2011, 10:32

There are build scripts in the main source directory.

Blight

22nd June 2011, 10:54

The first type is called "IMA4". Its an ADPCM codec, and ffdshow should support it.
The second type is not E-AC3, its a LAV Splitter internal type. Someone mislabeld it in some ZP profile.

Using 'ffdshow_rev3892_20110620_clsid' build from 2 days ago, ffdshow isn't able to connect to LAV Splitter for this audio stream (see attached screenshot).

Edit:
1. Is there an official sub-type for E-AC3?
2. Reading the changelog, there's indication of a programmable interface. Please include the headers in the ZIP/Installer.

HTPC-User

22nd June 2011, 13:01

hoborg

22nd June 2011, 13:03

Hi nevcairiel!

In MPC-HC one can switch to the next audio language and subtitle track by pressing a keyboard shortcut.
Is this also possible with your splitter? If it is possible, what do I have to do?

:thanks:

You can do this if you use FFDshow video decoder.
Very useful for control audio/subtitles streams switching, turn on/off subtitles, etc... by remote controler in all DS based players.

Gleb Egorych

22nd June 2011, 16:15

Edit:
1. Is there an official sub-type for E-AC3?
Look here (http://bugs.zoomplayer.com/bug.php?op=show&bugid=91) and here (http://msdn.microsoft.com/en-us/library/dd317599(v=vs.85).aspx).

RealSnoopyDog

23rd June 2011, 13:11

There are build scripts in the main source directory.

Uups, sorry, it was a bit late that day and i didn't recognize that there is a script. But now i was able to compile LAV - even with Visual Studio 2010 express. Thank you :)

RealSnoopyDog

24th June 2011, 10:49

Usually a MKV should have a Cue Point every few seconds, those cue points are used for the "fast seeking" in MPC-HC, if MakeMKV only puts them on chapters, i consider that a bug.

There seems to be a bug in the latest version:
http://www.makemkv.com/forum2/viewtopic.php?f=1&t=3634

One more question related to DTS Express: As far as i understand, DTS express is used for secondary audio tracks which are not designated to be played back standalone. They should mixed over the currently playing primary audio tracks. I found some newer Blu-Rays in my collection which have a DTS express secondary audio stream. You think it's hard to implement a mixing feature like this:When these new features are streamed in, your player will actually mix the audio portion of the “new” content with the audio already on the disc. The new audio is called Secondary Audio on Blu-ray Disc and can be up to 5.1 channels; on HD DVD this is called Sub Audio and can be in 2.0 channels. DTS uses a system called DTS Express to deliver Secondary or Sub Audio to your player. The DTS Express audio is mixed with the Primary Audio, giving you an intensely great interactive surround experience.

To ensure that the audio mix comes out perfectly, DTS-HD employs an innovative feature called Dynamic Mixing. When the movie or concert creators encode the Secondary or Sub audio, they include meta-data that will dynamically mix that audio with the primary audio. With Dynamic Mixing, you’ll continually have the correct sound balance — just as the producers intended.
Reference: http://www.dts.com/DTS_Audio_Formats/DTS-HD_Master_Audio/How_Does_DTS-HD_Work.aspx

But there are not many discs which come with this feature so i think the effort for implementing this is too high (?) and can be used better for other improvements. Or is the Arcsoft DLL already capable of this? For this reason, secondary audio streams should not be listed in the normal IAMStreamSelect togehter with the primary streams, once a mixing feature is available.

It's also working with DVBViewer (see screenshot) where i played back one of my blu-rays. But i manually re-create the contents of the audio menu and i have to append the secondary audio track to the end of the menu.

Blight

24th June 2011, 12:15

Gleb:
Thanks, I've updated the profile and closed the bug report :)

RealSnoopyDog

24th June 2011, 13:10

For this reason, secondary audio streams should not be listed in the normal IAMStreamSelect togehter with the primary streams, once a mixing feature is available.
OK, not a good idea from my side. It's good as it is right now. If the user selects a secondary audio stream from within IAMStreamSelect then the secondary stream can be played back standalone (as it is right now) or mixed together with the current primary stream (possible future). To go back to the previous state, just select the previous active stream.

I just changed the DVBViewer audio menu: If there are subaudio streams available, then it shows a menu entry at the bottom which is labeled "Secondary audio tracks...". This entry points to a sub menu where you can either select the secondary audio track(s) or "None" which switches back to the previous active audio track (see screenshot). I got it working like this with DVBViewer + LAV.

nevcairiel

25th June 2011, 07:38

Not all secondary streams are really required to be mixed onto the main stream, most of the time they contain the commentary for PiP streams - which you can mix, but the original audio would fade out significantly anyway, so can as well play it standalone.

RealSnoopyDog

25th June 2011, 09:25

You're right: Pip is secondary video which is mixed togehter with the primary video. The same with secondary audio. I found two Blu-Rays right now with secondary audio. In both cases the secondary audio track is silent most of the time because you should hear the original soundtrack with the lanuage that you selected at that time. Is it hard to realize some kind of mixing with DirectShow? You'll need another audio output pin (?)

simpleway

25th June 2011, 14:41

I tested it for mkv with real video (RV40), this video pin can't connect to grabest real video decorder

HTPC-User

25th June 2011, 16:17

You can do this if you use FFDshow video decoder.
Very useful for control audio/subtitles streams switching, turn on/off subtitles, etc... by remote controler in all DS based players.

Hi!

I am not quite sure if I can use this in my planned scenario: MPC-HC with ffdshow (for deinterlacing and avisynth scripts), LAV Filters/splitter and madVR.
In order to see subtitles I have to use the MPC-HC subtitle engine with madVR. (As far as I know ffdshow subtitles aren't supported by madVR).

So I still wonder if it is possible to switch to the next subtitle and audio track by pressing a key (and using Nevcairiel's cool splitter).

But perhaps I only misunderstood your suggestion.

:thanks:

fastplayer

25th June 2011, 16:20

(As far as I know ffdshow subtitles aren't supported by madVR).
They are supported. Always have been.

hoborg

26th June 2011, 17:35

Hi!

I am not quite sure if I can use this in my planned scenario: MPC-HC with ffdshow (for deinterlacing and avisynth scripts), LAV Filters/splitter and madVR.
In order to see subtitles I have to use the MPC-HC subtitle engine with madVR. (As far as I know ffdshow subtitles aren't supported by madVR).

So I still wonder if it is possible to switch to the next subtitle and audio track by pressing a key (and using Nevcairiel's cool splitter).

But perhaps I only misunderstood your suggestion.

:thanks:

FFDshow put subtitles inside the video, so they will apear in any renderer.
It is good to renable resize in ffdshow too to have the same subtitles undepended on video size.

HTPC-User

27th June 2011, 10:45

Hi!

FFDshow put subtitles inside the video, so they will apear in any renderer.
It is good to renable resize in ffdshow too to have the same subtitles undepended on video size.

Ok, I was wrong. I tried the ffdshow subs with EVR CP in MPC-HC and also the subtitles didn't work. So I think there is some kind of configuration problem on my side.
By the way I followed for example this tutorial http://www.homecinema-hd.com/ffdshow-video-subtitles_en.html but no subtitles. Any specific setting to be aware of? (Autoload subs in MPC-HC playback is off, subtitles in ffdshow are on, all transform filters in MPC are unchecked)

Anyway, as stated in several postings and threads the MPC-HC subtitle engine seems to be the better choice for subs than ffdshow and therefore I would like to use it.

That leads us to my original question :): Is it somehow possible to switch to the next audio and subtitle language by pressing a keyboard shortcut (and using MPC-HC's subtitle engine, LAVFilter/splitter, ffdshow for deinterlacing and scripts, madVR etc.)?

:thanks:

edison

27th June 2011, 15:40

Sometimes, the LAV Splitter detected some 59.94p mkv files as 59.92p.

http://www.gokuai.com/w/1bsbF0G29e4g4A0n/lav_59.940_59.917fps.jpg

and sometimes detected as 60.000p
http://www.gokuai.com/w/5L0mjw4zA3U1gRm0/lav_59.940_60.000fps.jpg

Selur

29th June 2011, 09:31

Trying to play a mp4 with multiple test subtitles (http://www.multiupload.com/4VDVDSAZ5B) in MPC-HC using LAV Filters, LAV CUVID and EVR CP.
Video works fine, but the subs don't. (they work fine if I use MPC-HC internal mp4 splitter)
-> is it a bug or just me overlooking something?

Cu Selur

adam777

29th June 2011, 12:27

Hello all,
I've read what seems relevant in the thread, but still could not understand the exact details regarding DTS-HD decoding, so if someone could make it simple for me it would be great.
I'm using LAV Splitter under Windows 7 x64 and my system's speaker setup is 5.1.
Also, I'm using ReClock's WASAPI exclusive as audio renderer.
What should I expect under the following:

1. Decoding a 7.1 DTS-HD file using LAV Audio (decoding quality (full? core?), output configuration).

2. Decoding a 7.1 DTS-HD file using Arcsoft decoder (the dll, same parameters).

3. Decoding a 5.1 DTS-HD file using LAV Audio (same).

3. Decoding a 5.1 DTS-HD file using Arcsoft decoder (same).

:thanks:
Adam

nevcairiel

29th June 2011, 13:52

adam777

29th June 2011, 14:11

LAV Audio alone (using ffmpeg) can only decode the Core, and if with "ArcSoft decoder" you mean using dtsdecoderdll.dll in LAV Audio then the following applies:

So -

1. 5.1 Core
2. Full 7.1 HD
3. 5.1 Core, same as 1)
4. Full 5.1 HD

OK, great.
Is there a combination that allows me to get something like full HD 5.1 in the second case?
Meaning, some sort of downmixing done by the arcsoft decoder that converts the 7.1 audio to 5.1, leaving it 24 bit etc.?

nevcairiel

29th June 2011, 14:14

No. But since its already decoded to PCM, you can use something like ffdshow for downmixing, without downsamplig.

LAV Audio will support (simple) downmixing eventually, but i don't know when i'll work on that.

adam777

29th June 2011, 14:15

Thanks again, it all makes sense now.

Oh, forgot to ask, what is the "correct" way of downmixing? extracting the core from the original stream or decoding at full resolution and let FFDShow handle the downmixing?

* EDIT *
I'll be a bit more specific.
Which way is generally considered "better" - utilizing the built-in 5.1 DTS core, thus receiving perfect sound matrix, but losing the details available in the MA stream.
Or decoding the stream in full resolution, maintaining all details, but losing (?) the sound mix due to the external downmixing.

On a side note, I think I've read something regarding an option to output the 7.1 MA stream at both 7.1,6.1 and 5.1 in a lossless fashion natively, which is obviously ideal.