LAV Filters - DirectShow Media Splitter and Decoders

[cncb]

I have some Blu-rays with 5.1 PCM audio that I want to send over HDMI to my AVR. I am able to do this successfully with LAVAudio in a player with WASAPI output. However, there is another Directshow player I want to use that doesn't support WASAPI output. Is the only way to do this by using ReClock or is there some other "trick" to get this to work? I really have no use for the other features of ReClock so I would like to minimize use of extraneous filters if possible. Thanks.

[nevcairiel]

I finished implementing the first hand-optimized pixel converter, in this case YUV 4:4:4 9/10bit -> AYUV, with dithering, in SSE2

Before: ~120fps at ~30% CPU usage
After: ~250fps at ~90% CPU usage

In both cases, the decoding is multi-threaded, but the conversion is not (Quad Core with HT, so 8 virtual cores)
The new code is fast enough to let the decoder run at nearly maximum speed. I call this success. :D

Its really painfully obvious now how SSE2 etc are really designed for image processing, the functions they offer just fit for what i need. :)

[SamuriHL]

Can't we have 250fps at ~30% CPU usage??? :D

[VipZ]

Nev, is there any chance to revert this change, e6bc58c658cd ?

Allowing the MS decoder to connect is pointless IMO as it can't decode VC1 in EVO with your splitter anyway, while LAV Video has no issues. This just makes VC1 more painful to make a solution around.

[v0lt]

FLICs do not play.
Example (http://samples.mplayerhq.hu/fli-flc/HWdopple.flc)

[madshi]

I finished implementing the first hand-optimized pixel converter, in this case YUV 4:4:4 9/10bit -> AYUV, with dithering, in SSE2

Before: ~120fps at ~30% CPU usage
After: ~250fps at ~90% CPU usage
Nice! JFMI, what kind of dithering have you ended up with? Ordered dithering?

Allowing the MS decoder to connect is pointless IMO as it can't decode VC1 in EVO with your splitter anyway
If that is the case then why not fixing this problem? Decoding VC1 in EVO works with the Haali Media Splitter, so it should be possible to make it work with the LAV Splitter, too.

[nevcairiel]

Nice! JFMI, what kind of dithering have you ended up with? Ordered dithering?

Yes, pretty standard ordered dithering with a Bayes matrix - 8x8, because it fits so perfectly on the SSE2 logic (8 16-bit pixels fit into a SSE2 register)


Nev, is there any chance to revert this change, e6bc58c658cd ?

Allowing the MS decoder to connect is pointless IMO as it can't decode VC1 in EVO with your splitter anyway, while LAV Video has no issues. This just makes VC1 more painful to make a solution around.

Decoding works for me on the few sample EVO clips people gave me. Needs FTC when using EVR-CP though, but is fine on madVR.
Thats what the change was all about, making it work. :)

In any case, i won't break a decoder on purpose. If people don't want to use it, their player should somehow manage that for them.

[Midzuki]

To whom this may concern :) , MPG-in-MP4 started to work after I un-registered MPC-HC's MP4Splitter.ax and all :devil: Haali DLLs. I don't know why, but the CPU-usage is reduced by 50% when I use LAV's MP4 splitter together with Cyberlink's MPEG decoder :) :) :)

P.S.: :thanks: again.

[VipZ]

that is the case then why not fixing this problem? Decoding VC1 in EVO works with the Haali Media Splitter, so it should be possible to make it work with the LAV Splitter, too.

The main reason, is that no one wants to put effort to fix EVO anymore as its a dead format, I am just trying to keep LAV working perfectly with EVO as it was.

Decoding works for me on the few sample EVO clips people gave me. Needs FTC when using EVR-CP though, but is fine on madVR.
Thats what the change was all about, making it work. :)

In any case, i won't break a decoder on purpose. If people don't want to use it, their player should somehow manage that for them.

I was under the impression that FTC on VC1 was never needed with LAV Spitter so never tried this.

The main reason I like that other filters never connected was that you could use ffdshow as a higher merit to handle any VC1 content via the wmv9 decoder which works very well, but still having LAV Video to solely deal with VC1 in EVO's. This setup is completely hardware independent and requires no real setup so anyone can easily achieve full VC1 playback.

Personally I am just reverting this commit in my builds to make it easier.

[nevcairiel]

I was under the impression that FTC on VC1 was never needed with LAV Spitter so never tried this.

It is for EVOs because the MS decoder does not produce 100% smooth timestamps because of the forced pulldown.
But it works perfectly otherwise.

[STaRGaZeR]

I finished implementing the first hand-optimized pixel converter, in this case YUV 4:4:4 9/10bit -> AYUV, with dithering, in SSE2

Before: ~120fps at ~30% CPU usage
After: ~250fps at ~90% CPU usage

In both cases, the decoding is multi-threaded, but the conversion is not (Quad Core with HT, so 8 virtual cores)
The new code is fast enough to let the decoder run at nearly maximum speed. I call this success. :D

Its really painfully obvious now how SSE2 etc are really designed for image processing, the functions they offer just fit for what i need. :)

http://www.xtremesystems.org/forums/images/smilies/clap.gifhttp://www.xtremesystems.org/forums/images/smilies/clap.gifhttp://www.xtremesystems.org/forums/images/smilies/clap.gif

Any chance to see the speed improvement of the conversion itself? :D

[nevcairiel]

Any chance to see the speed improvement of the conversion itself? :D

The big improvement is only visible in release builds, because debug builds do not fully optimize the intrinsics (for debugging purposes, its copies the values back into memory after every step, so the debugger can see them)

But in release builds, the debug timing logic is turned off..
Lets see if i can maybe activate the intrinsic optimization.

[VipZ]

It is for EVOs because the MS decoder does not produce 100% smooth timestamps because of the forced pulldown.
But it works perfectly otherwise.

Is there any harm in leaving FTC on all the time with your filters?

Also that commit, was this pretty much just against EVO or MPEG in general?

[nevcairiel]

Also that commit, was this pretty much just against EVO or MPEG in general?

No idea what you mean with "against" - the media type of VC-1 in EVO was just wrong, and i fixed it.

[VipZ]

No idea what you mean with "against" - the media type of VC-1 in EVO was just wrong, and i fixed it.

Let me try rephrase :)

The change in that commit to fix the VC1 media type, does this affect EVO files only, or mpeg in general.

[nevcairiel]

All mpeg-ps files, although i'm unsure if VC-1 exists in normal MPEG-PS files..

[VipZ]

Thanks for the info, was curious if I decided to keep the commit reverted in my builds what the side affects could be.

[nevcairiel]

I still don't get what thats good for. Just set LAV Video to be your preferred decoder, and it will be used.
If you want different decoders based on the source medium, maybe you should switch to a player that allows a fine-grained control like that. ;)

[VipZ]

LAV Video is my preferred decoder and already kicked ffdshow out with the exception for VC1 as it handles interlaced content perfectly. I am just torn between either VC1 interlaced or VC1 in EVO being more important, Before that commit I didn't have to decide as it just worked :)

With my Playback Kit I make, I want it to be universal in playback ability without needing to change settings based on content.

I am open to another suggestion to play back VC1 universally without any special requirements.

[nevcairiel]

If ffmpeg doesn't implement VC-1 "soon" (lets say, this year), i might be persuaded to implement the MS wmv9 decoder. But for now, alot of other things to do.

[VipZ]

Cool, glad to know you will consider it if it's becomes necessary :)

But agreed while VC1 is the only real annoyance, its not big enough to warrant putting on top of any list. The combo of LAV Video and ffdshow for VC1 will work for the time being.

[lauhangwoo]

When I'm using "Repeat forever" option in MPC-HC, this flv having proper video/audio in the 1st playthrough, but no video/audio in the 2nd playthrough.
http://www.geocities.co.jp/lauhangwoo/etc/sample2.zip

[nautilus7]

Anyone else having problems with the Arcsoft DTS decoder? I have placed the dll in the LAV directory but in status keeps saying "dca" when i play a master audio track.

[VipZ]

Nev, it seems the Read AR from stream option no longer works with h264/MKV's.

[nevcairiel]

Nev, it seems the Read AR from stream option no longer works with h264/MKV's.

The option works just fine, it looks like ffmpeg fails to read the AR from the H264 header...
Are you sure it worked before? :D

[Midzuki]

Anyone else having problems with the Arcsoft DTS decoder? I have placed the dll in the LAV directory but in status keeps saying "dca" when i play a master audio track.

What version of the DTSdecoder DLL do you have? :confused:
Mine is 1.1.0.1
Also, are you playing it thru LAV Splitter, or thru something else?

[Budtz]

Seems ffdshow added support for decoding H.264 10-bit 4:2:0 and 4:4:4 since r3954 so there seems not much difference between lav video and ffdshow other than ffdshow has lots more post processing and other configurations options. Or am i missing something?

Lav of cause is much easer to setup and use.

[VipZ]

The option works just fine, it looks like ffmpeg fails to read the AR from the H264 header...
Are you sure it worked before? :D

Yep, It used to work perfectly before. There have been so many changes recently, I have no clue when it stopped working, I can say it was a build older than Sunday last week.

This is just some guesses, but I would maybe think,
e89e21b0e39d-Ensure the Aspect Ratio is not 0
18505222dd93-Add support for the VIDEOINFOHEADER output format type.

I will try test old revisions and see which one broke it. EDIT, its between 06-08 and 14-08, so many changes :)

[nevcairiel]

Its not any change i did, ffmpeg just doesn't report a AR anymore. It works on some MPEG-2 streams where i knew the AR changes mid-stream, but on the MKV it just reports no AR at all.
If you could figure out in which of my uploaded versions it first broke, i could narrow down the range of ffmpeg commits, otherwise its insane to find it.

Edit:
I just tested "LAVFilter-video1.zip", the first version ever uploaded, and on the MKV i have it doesn't work. Yet, LAV CUVID somehow manages to figure it out.
I'm confused!

Edit2:
Or maybe it doesn't? It seems to go to the same AR as the file dimensions, this was 720x480, and it claims its 3:2, which matches exactly. Maybe i'm testing with a wrong file. :P
I did re-test my MPEG2, and i'm sure that the option itself works just fine.

[VipZ]

This is the sample stream is 4:3, but has a 16:9 MKV AR Flag, http://www.mediafire.com/?t1i1fqcj6qhwb4s

Under default conditions, this will display as 4:3, but if that use stream AR option was unticked, it would then take the 16:9 AR flag, a build I did on the 6th is working. I am battling with git to try get some intermediate builds.

EDIT: Took a guess and reverted this commit, e89e21b0e39d-Ensure the Aspect Ratio is not 0. It works again after this.

Yep its fixed :) Its not as dynamic as before, but that's not an issue. (Can switch to stream AR, but not back to flag AR)

[nevcairiel]

Oh, i know whats going on, i think.
Its really the other way around. Disabling that option doesn't really work. And i thought enabling it didn't work. :)

Should hopefully be fixed.

[Cudo]

Hi,

In MPC-HC x86 on Win 7 x64, no matter which decoders / splitter I use, I always have to set 250 ms audio delay to fix an A/V sync issue when playing x264 .mkv video files which contain DTS, DTS-HD or DTS-HD MA audio. However, unlike AC3Filter and ffdshow audio, LAV Audio decoder has no built-in option to set a delay. Is there anything I can do to work my way around this? My sound setup is analog 5.1 surround via onboard Realtek HD Audio. Any help would be greatly appreciated.

[nevcairiel]

MPC-HC itself has a feature to control audio delay.

[Cudo]

Anyone else having problems with the Arcsoft DTS decoder? I have placed the dll in the LAV directory but in status keeps saying "dca" when i play a master audio track.
No, my Arcsoft dll is 1.1.0.8 and LAV Audio status says it's DTS-HD MA.

[Cudo]

MPC-HC itself has a feature to control audio delay.Woow, that was fast! :D I can't see farther than the tip of my nose I guess lol. Thanks an octillion. :)

[jmone]

Nevcairiel - did you get a change to try these VC-1 samples?
Thanks
Nathan

Thanks.

Did some testing with an EVO from "Heroes, S1E1"

MediaInfo reports it as:

Video
ID : 253 (0xFD)-85 (0x55)
Format : VC-1
Format profile : AP@L3
Duration : 53mn 17s
Bit rate : 18.6 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate : 23.976 fps
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Scan order : 2:3 Pulldown
Compression mode : Lossy
Bits/(Pixel*Frame) : 0.374
Stream size : 6.92 GiB (91%)


When I play it with LAVSplitter / LAVCUVID / madVR the movie is reported as 59.94 and I get presentation glitches and it parts it stutters badly.

The same file when remuxed to M2TS plays at 23.976 and while it does not have the presentation glitches it too does not look very good with panning and will stutter on these scenes badly.

Samples:
Original EVO: http://www.megaupload.com/?d=BOC0RSXW
Remuxed M2TS: http://www.megaupload.com/?d=H5USKBUW

[nevcairiel]

VC-1 EVO being reported as 59.94 is normal, there is nothing really to be done. It looks like a interlaced stream at 60i, but when decoded only outputs 24 progressive frames per second. It just doesn't know any better.
The whole concept of forced pulldown is weird to handle, especially if you want auto-refresh rate switching to work. I'm so glad HDDVD is not the format that prevailed.

[nautilus7]

What version of the DTSdecoder DLL do you have? :confused:
Mine is 1.1.0.1
Also, are you playing it thru LAV Splitter, or thru something else?

No, my Arcsoft dll is 1.1.0.8 and LAV Audio status says it's DTS-HD MA.

Well, that's it. Problem is with v1.1.0.0. I tested with v1.1.0.8 which I also have and it's working fine.

[jmone]

VC-1 EVO being reported as 59.94 is normal, there is nothing really to be done. It looks like a interlaced stream at 60i, but when decoded only outputs 24 progressive frames per second. It just doesn't know any better.
The whole concept of forced pulldown is weird to handle, especially if you want auto-refresh rate switching to work. I'm so glad HDDVD is not the format that prevailed.

Thanks - for the explanation.

What I'm trying to work out is why my other EVO and remuxed M2TS play fine and are nice and smooth but both the sample EVO and M2TS looks terrible in comparison with all sorts of judder.

Check at the stutter around the 45-sec mark on the EVO and even the M2TS remux looks terrible to me.

The only thing I can think of is the rest have been either Movie Content (original 24fps), or Music Video (original 60i). I've just got to the first set of TV Series (?? on what it was originally shot at prior to encoding).

[SamuriHL]

VC-1 EVO being reported as 59.94 is normal, there is nothing really to be done. It looks like a interlaced stream at 60i, but when decoded only outputs 24 progressive frames per second. It just doesn't know any better.
The whole concept of forced pulldown is weird to handle, especially if you want auto-refresh rate switching to work. I'm so glad HDDVD is not the format that prevailed.

ARGH!!! Yes, you REALLY should be using eac3to to rip out that video!!!!! It'll deal with the pulldown correctly. The EVO's themselves, yea, I hear what you're saying. It's just VC1 that's the issue, however.

[SamuriHL]

Thanks - for the explanation.

What I'm trying to work out is why my other EVO and remuxed M2TS play fine and are nice and smooth but both the sample EVO and M2TS looks terrible in comparison with all sorts of judder.

Check at the stutter around the 45-sec mark on the EVO and even the M2TS remux looks terrible to me.

The only thing I can think of is the rest have been either Movie Content (original 24fps), or Music Video (original 60i). I've just got to the first set of TV Series (?? on what it was originally shot at prior to encoding).

Because you had the pulldown flags removed. ;) (or didn't on the one that isn't smooth)

[jmone]

The EVO sample is a RAW EVO straight from the HD-DVD. The M2TS sample is created from the streams being extract by eac3to with the pulldown removed and then remuxed with tsmuxer (hence it is now reported as 23.976). I've re-check with other EVO and remuxed M2TS versions and they look fine (including other TV show - Battlestar Galaxtica), so maybe it is just something bad with the Heroes encode.... does anyone see the stuttering issues on these samples or is it just me?

[jmone]

FYI - here is part of the eac3to log incase I missed something:
eac3to v3.24
command line: "x:\transcodes\eac3to\eac3to.exe" "D:" 3) 1: "x:\Output\Genesis\Chapters_1.txt" 2: "x:\Output\Genesis\Video_2.*" 3: "x:\Output\Genesis\Audio_3_English.W64" -progressnumbers -LOG="x:\Output\Genesis\LOGS\eac3to_PASS3_LOG.LOG"
------------------------------------------------------------------------------
EVO, 1 video track, 1 audio track, 2 subtitle tracks, 0:53:17
1: Chapters, 4 chapters with names
2: VC-1, 1080p24 /1.001 (16:9) with pulldown flags
3: E-AC3, English, 5.1 channels, 1536kbps, 48kHz, dialnorm: -27dB
4: Subtitle (DVD), English, "SDH"
5: Subtitle (DVD), French
Creating file "x:\Output\Genesis\Chapters_1.txt"...
[a03] The Nero decoder doesn't seem to work, will use libav instead.
[v02] Extracting video track number 2...
[a03] Extracting audio track number 3...
[v02] Writing new framerate "24fps /1.001" to bitstream.
[a03] Removing E-AC3 dialog normalization...
[v02] Removing VC-1 pulldown...
[a03] Decoding with libav/ffmpeg...
[a03] Remapping channels...
[a03] Reducing depth from 64 to 24 bits...
[a03] Writing W64...
[a03] Creating file "x:\Output\Genesis\Audio_3_English.W64"...
[v02] Creating file "x:\Output\Genesis\Video_2.vc1"...
[a03] Clipping detected, a 2nd pass will be necessary. <WARNING>
[a03] Starting 2nd pass...
[a03] Extracting audio track number 3...
[a03] Removing E-AC3 dialog normalization...
[a03] Decoding with libav/ffmpeg...
[a03] Remapping channels...
[a03] Reducing depth from 64 to 24 bits...
[a03] Writing W64...
[a03] Applying -0.36dB gain...
[a03] Creating file "x:\Output\Genesis\Audio_3_English.W64"...
Video track 2 contains 76658 frames.
eac3to processing took 23 minutes, 19 seconds.
Done.

[SamuriHL]

I'm home alone tomorrow for most of the day and I own Heroes season 1 on HD DVD so I can mess with it myself and see what I can come up with.

[jmone]

Thanks

[FlashGordon]

I can't get LAV Video to work with the MS DVD Navigator even though it plays individual VOB files just fine. Am I the only one experiencing this or is the navigator just not supported yet?

[PepsiLee2001]

Dear Sir,

I find two problems about LAV Splitter & audio decoder.

1. it seems that can't split TrueHD & DTS-HD MA 5.1 192K/24bit sound track currently. (from http://www.eastwindimport.com/product-info.asp?ProductId=647)
2. Dobly Digital Plus can't bitstream to AV Receiver. (from Super HiVi Cast)

[dann23]

wmp12 doesn't show subtitles when using lav filters. Is this normal or is it a bug?

[ageback]

How to translate LAV Filters' Setting dialogs to other languages?

[VipZ]

wmp12 doesn't show subtitles when using lav filters. Is this normal or is it a bug?

You will need to use ffdshow RAW for this or vsfilter, but haven't tried to use myself within WMP12



Nev, sorry to bug you again :)

Now that the MKV AR option is fixed, WMV AR is broken again.

Both MKV and WMV work if the following commits are reverted,
520504295eb6-Fix disabling of the Stream AR option
e89e21b0e39d-Ensure the Aspect Ratio is not 0

Also with the recent changes to promote using the MS VC1 decoder, which I am also now using. The MS VC1 Decoder within Windows XP doesn't connect to LAV Splittler but has no issues connecting to the MPC MPEG splitter, I don't care for XP and just occasional test stuff within my Virtual Machine, but I would think this would be an easy enough thing to fix. If you need me to do anything let me know.

Info on the connection,
Filter : Mpeg Splitter (low merit) - CLSID : {DC257063-045F-4BE2-BD5B-E12279C464F0}

- Connected to:

CLSID: {94297043-BD82-4DFD-B0DE-8177739C6D20}
Filter: WMVideo Decoder DMO
Pin: in0

- Connection media type:

Video: WVC1 1920x1080 29.97fps

AM_MEDIA_TYPE:
majortype: MEDIATYPE_Video {73646976-0000-0010-8000-00AA00389B71}
subtype: Unknown GUID Name {31435657-0000-0010-8000-00AA00389B71}
formattype: FORMAT_VideoInfo2 {F72A76A0-EB0A-11D0-ACE4-0000C0CC16BA}
bFixedSizeSamples: 1
bTemporalCompression: 0
lSampleSize: 1
cbFormat: 149

VIDEOINFOHEADER:
rcSource: (0,0)-(0,0)
rcTarget: (0,0)-(0,0)
dwBitRate: 0
dwBitErrorRate: 0
AvgTimePerFrame: 333666

VIDEOINFOHEADER2:
dwInterlaceFlags: 0x00000000
dwCopyProtectFlags: 0x00000000
dwPictAspectRatioX: 16
dwPictAspectRatioY: 9
dwControlFlags: 0x00000000
dwReserved2: 0x00000000

BITMAPINFOHEADER:
biSize: 40
biWidth: 1920
biHeight: 1080
biPlanes: 0
biBitCount: 0
biCompression: WVC1
biSizeImage: 0
biXPelsPerMeter: 0
biYPelsPerMeter: 0
biClrUsed: 0
biClrImportant: 0

pbFormat:
0000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
0010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
0020: 00 00 00 00 00 00 00 00 62 17 05 00 00 00 00 00 ........b.......
0030: 00 00 00 00 00 00 00 00 10 00 00 00 09 00 00 00 ................
0040: 00 00 00 00 00 00 00 00 28 00 00 00 80 07 00 00 ........(...€...
0050: 38 04 00 00 00 00 00 00 57 56 43 31 00 00 00 00 8.......WVC1....
0060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
0070:|00 00 00 01 0f db fe 3b f2 1b ca 3b f8 86 f1 80 .....Ûþ;ò.Ê;ø†ñ€
0080: ca 02 02 03 09 a5 3e bf 07 fc 00 00 01 0e 1a df Ê....¥>¿.ü.....ß
0090: fc ef c8 6c 40 üïÈl@

- Enumerated media type 0:

Set as the current media type