ffdshow tryouts project: Discussion & Development

[STaRGaZeR]

Guys, I'm getting a buttload of errors when compiling ffmpegmt for x64, in pthread.c:

http://i32.tinypic.com/2mp0vhg.png

They don't make any sense. No errors when compiling for x86, like it should be. Any suggestions?

[clsid]

You need to put pthread headers/libs in your MinGW64 installation.

[haruhiko_yamagata]

I disagree with you on this one.
according to this MSDN page :
http://msdn.microsoft.com/en-us/library/dd377618%28VS.85%29.aspx
my previous explanation seems to be good.
so this is not a bug in the splitter which correctly does its job but in ffdshow.
set the rectangle infos (rcSource) in the buffer infos (BITMAPINFOHEADER) in the splitter could break other decoders which rely on the buffer infos to decode the stream.
OK, then please commit the patch.

[MatMaul]

OK, then please commit the patch.
done with a little modification to also handle theora cases.

[STaRGaZeR]

You need to put pthread headers/libs in your MinGW64 installation.

Done, but now I get this:

http://thumbnails9.imagebam.com/4194/fa59f441931332.gif (http://www.imagebam.com/image/fa59f441931332)

[clsid]

I have noticed a few issues with the patch.

1) The cropping does not work properly when using the Overlay Mixer. The previously blurry parts now show whatever is visible in the background at the time the playback starts.
2) The "Output size" value of the OSD reports the original size.

Edit: it also doesn't work with VMR-7.


@STaRGaZeR, maybe you need a different build of pthread.a. Sorry, I don't have time to help you further.

[STaRGaZeR]

The only one I know of is the one present here: http://sourceforge.net/projects/mingw-w64/files/

Which I've already tried with the same results :(

[mark0077]

Guys, been watching alot of old content recently that needs to be de-interlaced. I have found moving to hardware de-interlacing is giving me alot nicer de-interlacing than yadif so I am happy enough for the moment outputting nv12 from ffdshow.

My problem again comes back to DVD's that are badly marked as coming from interlaced sources (most of my DVDs at least just need simple weave to 24fps). My question is (as there isn't a good software fix /detector for this yet), is there a way to change the interlaced flag of each frame with one setting in ffdshow. I need this because for those disks that are badly marked, nvidia de-interlaces so I get 48fps instead of 24, aswell as slight artifacts when the nvidia de-interlacer messes with the frames. I would much rather the 24 original progressive frames that are there.

I have a script that I made months ago that is called from xbmc when I start a movie / tv show, which looks up a text file containing content that 'really' needs to be de-interlaced. Previously using software de-interlacing, I made this script write the ffdshow de-interlacing setting in the registry to on or off, (yadif on or off), but now that I am outputting the interlaced flag to the hardware de-interlacer, this main ffdshow de-interlacing option no longer can work. I am in search of another setting that will let me force each frame to be detected as interlaced or progressive instead of Auto, and effect everthing it should within ffdshow.

I suppose I could change the "Method:" in the "Output" tab in ffdshow, to "Force Weave" or "Force Bob", but would this mean that other processes within ffdshow would be effected?? Mainly what I ask here is, is good yv12 -> nv12 conversion dependent on proper interlaced flag detection.... because I see conversion to YV12 -> YUY2 needs this "Interlace or progressive" option to be set correctly in the "RGB Conversion" area, but I am unsure if it also needs to be set for YV12 -> NV12 only. If not I am hoping I can just find the registry setting for "Output" -> "Method", and change this using my script to have some sort of automation.

EDIT: Actually looking closely at some scenes, yadif actually does a much better job than outputting nv12 to nvidia... I wonder do I need to set something special to use NVidias best de-interlacing method. Control panel options are limited to one setting "Use inverse telicine" which I have enabled. Doing a frame step on some scenes that show it off, and nvidia does an absolutely terrible job at some scenes outputting both yuy2 and nv12... I'm confused now... I am in Windows 7 with latest 186.24 drivers.

[Mercury_22]

E-AC3 (Dolby Digital Plus) channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - SIDE LEFT wrong mapped to REAR LEFT

= No Side Channels

Test 7.1 E-AC3 (Dolby Digital Plus) (http://sharebee.com/69772daa)

Dolby TrueHD channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - REAR RIGHT wrong mapped to SIDE RIGHT
3 - SIDE LEFT wrong mapped to REAR LEFT
4 - REAR LEFT wrong mapped to SIDE LEFT

Test 7.1 Dolby TrueHD (http://www.megaupload.com/?d=7P9P81CW)

[mark0077]

E-AC3 (Dolby Digital Plus) channels are wrong mapped :
1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - SIDE LEFT wrong mapped to REAR LEFT
= No Side Channels
Test file http://sharebee.com/69772daa

Dolby TrueHD channels are wrong mapped :
1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - REAR RIGHT wrong mapped to SIDE RIGHT
3 - SIDE LEFT wrong mapped to REAR LEFT
4 - REAR LEFT wrong mapped to SIDE LEFT

Test file http://www.megaupload.com/?d=7P9P81CW

I'm hoping this is the key to my HDAV Deluxe problems, rear right and rear left sound like they are getting blurred together with ac3 5.1 surround sound tests. BUT it also happens for wma 5.1 files. Also using reclock to output in wasapi exclusive mode this rear blurring effect seems to disappear.

If someone confirms this is a bug in mpc-hc, could it apply to all rear outputs, and only in directsound mode? I get correct ac3 5.1 channels using wasapi out but not directsound.

[Mercury_22]

I'm hoping this is the key to my HDAV Deluxe problems, rear right and rear left sound like they are getting blurred together with ac3 5.1 surround sound tests. BUT it also happens for wma 5.1 files. Also using reclock to output in wasapi exclusive mode this rear blurring effect seems to disappear.

If someone confirms this is a bug in mpc-hc, could it apply to all rear outputs, and only in directsound mode? I get correct ac3 5.1 channels using wasapi out but not directsound.

ac3 5.1 surround sound test (posted by me on MPC-HC's thread) works fine with FFD ! (it's not working with MPC-HC's INTERNAL filters ! )

P.S. I might be wrong but I think you meant to post this on MPC-HC's thread not on FFD's ? ( MPC's INTERNAL filters wrong mapped channels are different from FFD's wrong mapped channels)

[travolter]

In the codec tab I set all decoders to use libavcodec.

Ev3rything OK .. I play WMV/VC-1/MKV/MPG/MOV/MP4 ok... but when I load a XVID/DIVX.. the performance is terribly bad.. I lose frames when I use same filters that are working ok in 1080VC-1 videos..

I have a Quad core CPU and I notice that only one core is really boosted..


There is any way to fix that?

Maybe installing some compatible Xvid/divx dshow codec?


EDIT.- I solved the problem installing Haali splitter.. and enabling the option "decoder priority boost"
(any fix if you dont want to have Haali installed?)

[MatMaul]

I have noticed a few issues with the patch.

1) The cropping does not work properly when using the Overlay Mixer. The previously blurry parts now show whatever is visible in the background at the time the playback starts.
2) The "Output size" value of the OSD reports the original size.

Edit: it also doesn't work with VMR-7.
... microsoft seems to do not follow their own recommendation in their old renderers...
same "problem" appears with mpc internal decoder.
I will not try to fix that because I don't really care about those old renderers (and I have no idea how to fix that without doing a real crop of the output video buffer).

[clsid]

Considering that in absolute numbers, probably more people use the old renderers than the new ones, a fix would be very welcome. If that would require ffdshow to do the cropping instead of the renderer, then so be it. If you don't want to do it, then maybe Haruhiko can once he has some time.

[Mercury_22]

LPCM channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to SIDE LEFT
2 - SIDE LEFT wrong mapped to LFE
3 - LFE wrong mapped to SIDE RIGHT

Test 7.1 LPCM (http://www.megaupload.com/?d=0HM2JCEP)

[Amour]

Today I did the monthly Windows Update. (Vista)
Result ?
I got big sound issues: there is a gap of 2 seconds between sound and video.
I will try to re-install ffdshow.

[edit]
re-installing ffdshow doesn't work!
I think there is a big issue with my computer now, because I can't watch videos normally anymore. :(

[Amour]

I tested different players, and the issue only came with MPC-HC.
Sorry, it might not be an ffdshow issue after all.
I just wonder if it is related to this:
http://blogs.technet.com/msrc/archive/2009/07/08/july-2009-advance-notification.aspx

[edit]
I fixed it!
I manually deleted "HKEY_CURRENT_USER\Software\Gabest" and it did the trick. :)
Sorry for bothering you with something not related to ffdshow.

[liquidskin76]

Hi,

Just after an update regarding DTS-HD and TrueHD/DTS-HD passthrough.

Is there any new news on DTS-HD decoding (not DTS core passthrough)? Is this being developed? I read somewhere that it was pretty close however not sure if true?

Regarding TrueHD/DTS-HD passthrough (bitstreaming), am i right in thinking that as long as the source files are decypted (ie. using AnyDVD HD), PAP does not play a part and the passthrough is theoretically possible? Is this being developed for FFDShow to work alongside AC3/DTS passthrough (please god, say yes!!!). I believe SlySoft are hoping to implement bitstreaming of HD audio over an HDMI port with their new upcoming SlyPlayer.

Cheers

[Snowknight26]

If whatever you're passing TrueHD through to can decode TrueHD, then it can also decode LPCM. So why wouldn't you just have ffdshow decode TrueHD?

[liquidskin76]

If whatever you're passing TrueHD through to can decode TrueHD, then it can also decode LPCM. So why wouldn't you just have ffdshow decode TrueHD?

I know what you mean, it's just that i'd love to see TrueHD and DTS-HD light up on the av receiver!!

In the end i geuss i'd be happy with just DTS-HD decoding (as well as the TrueHD decoding already in place).

I've tried bitstreaming with the ASUS Xonar HDAV 1.3 and Arcsoft TMT3 however it's a mess of a setup!

Cheers

[Mercury_22]

Is it normal to have DTS SIDE RIGHT mapped to REAR RIGHT and SIDE LEFT mapped to REAR LEFT (is it DTS-HD HR or MA working)?

Test 7.1 DTS-HD HR (http://sharebee.com/abf92989)

[mark0077]

Mercury, because I have had trouble with both rear outputs with my soundcard and setup (some form of fuzzy mixing between the two) you have sparked my curiosity. I post here because I notice in ffdshow, in the mixer section, the "5 channels" output configuration uses Back L and Back R, where I would have always assumed the rear channels of a 5.1 track would be Side L and Side R. I have my rear speakers setup as per dolby recommendations, at the side, slightly to the rear.

So confused now as to what each piece of the chain intends, the original track, the software like ffdshow, and finally the drivers.

[Thunderbolt8]

whats that ffmpeg page again where I can also make feature requests?

[Mercury_22]

Mercury, because I have had trouble with both rear outputs with my soundcard and setup (some form of fuzzy mixing between the two) you have sparked my curiosity. I post here because I notice in ffdshow, in the mixer section, the "5 channels" output configuration uses Back L and Back R, where I would have always assumed the rear channels of a 5.1 track would be Side L and Side R. I have my rear speakers setup as per dolby recommendations, at the side, slightly to the rear.

So confused now as to what each piece of the chain intends, the original track, the software like ffdshow, and finally the drivers.

I think FFD'S settings are correct I don't know why SOME drivers are using SIDE instead of REAR that's why I recommend you to use 7.1 in windows sound panel even with 5.1 speakers (not so many tracks with 7.1) and you can use window's audio control panel to remap side to rear & front (just uncheck "side pair" in Vista's or Windows 7's "configure") if you have any 7.1 soundtrack

[albain]

I didn't test the sample yet, do we have a problem here on channels mapping ? It may be a problem with the channel mapping between ffmpeg and FFDShow. FFMpeg has recently integrated a channel remapper. I have fixed this already for some formats but they may have remap channels for other formats since.

Concerning DTS-HD decoding (MA or not), we depend on FFMpeg project.
However, DTS-HD format is much more complicated than DolbyTrueHD to decode, and of course there is no white paper on it.

About HDMI passthrough for DTS-HD/TrueHD/LPCM, I don't have the equipment (nor the OS) to test it : is this working for any of those ?
I guess that the work to do on FFDShow would be to encapsulate the decoded stream into LPCM, especially if this is decoded TrueHD or decoded DTS-HD

I know that Microsoft (on windows 7 only) brought some modifications to the directshow structures (to WAVEFORMATEXTENSIBLE) in order to deal with this.

[Peekstra]

Although ffdshow functions fine with most DVD menu's still some of them don't work as they should.

For example, the msmpeg decoder works fine here as it only shows the title where the mouse pointer hovers above it. But with ffdshow navigation is impossible because you can't see which movie has been selected.

Using Zoom Player, MS DVD navigator & ffdshow 3020.

[Mercury_22]

I didn't test the sample yet, do we have a problem here on channels mapping ? It may be a problem with the channel mapping between ffmpeg and FFDShow. FFMpeg has recently integrated a channel remapper. I have fixed this already for some formats but they may have remap channels for other formats since.

Concerning DTS-HD decoding (MA or not), we depend on FFMpeg project.
However, DTS-HD format is much more complicated than DolbyTrueHD to decode, and of course there is no white paper on it.

About HDMI passthrough for DTS-HD/TrueHD/LPCM, I don't have the equipment (nor the OS) to test it : is this working for any of those ?
I guess that the work to do on FFDShow would be to encapsulate the decoded stream into LPCM, especially if this is decoded TrueHD or decoded DTS-HD

I know that Microsoft (on windows 7 only) brought some modifications to the directshow structures (to WAVEFORMATEXTENSIBLE) in order to deal with this.

Yes we have :

E-AC3 (Dolby Digital Plus) channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - SIDE LEFT wrong mapped to REAR LEFT

= No Side Channels

Test 7.1 E-AC3 (Dolby Digital Plus) (http://sharebee.com/69772daa)

Dolby TrueHD channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - REAR RIGHT wrong mapped to SIDE RIGHT
3 - SIDE LEFT wrong mapped to REAR LEFT
4 - REAR LEFT wrong mapped to SIDE LEFT

Test 7.1 Dolby TrueHD (http://www.megaupload.com/?d=7P9P81CW)

LPCM channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to SIDE LEFT
2 - SIDE LEFT wrong mapped to LFE
3 - LFE wrong mapped to SIDE RIGHT

Test 7.1 LPCM (http://www.megaupload.com/?d=0HM2JCEP)

Is it normal to have DTS SIDE RIGHT mapped to REAR RIGHT and SIDE LEFT mapped to REAR LEFT (is it DTS-HD HR or MA working)?

Test 7.1 DTS-HD HR (http://sharebee.com/abf92989)

All tested on a 7.1 config, cause the 5.1 config in windows sound scheme has most of the time (depending of the driver) SIDE instead of REAR channels.

I'm sure they are wrong mapped cause the same files with other filters (even MPC-HC's for TrueHD) are playing the correct channel on my system

Although the last one, DTS-HD I think it's "normal" do to no DTS-HD HR or MA decoding, I guess

[tetsuo55]

I posted about correct channel order here:
Please use this information when fixing the problem

http://forum.doom9.org/showthread.php?p=1306597#post1306597

[Octo-puss]

I've got this weird clip with audio being reported as:

ID : 1
Format : ADPCM
Format profile : U-Law
Codec ID : ulaw
Duration : 1mn 43s
Bit rate mode : Constant
Bit rate : 256 Kbps
Channel(s) : 1 channel
Sampling rate : 32.0 KHz
Resolution : 16 bits


I can't get it to play at all. Is it supported in FFDS at all?

[poisondeathray]

I've got this weird clip with audio being reported as:

ID : 1
Format : ADPCM
Format profile : U-Law
Codec ID : ulaw
Duration : 1mn 43s
Bit rate mode : Constant
Bit rate : 256 Kbps
Channel(s) : 1 channel
Sampling rate : 32.0 KHz
Resolution : 16 bits


I can't get it to play at all. Is it supported in FFDS at all?

Did you try enabling different adpcm choices in the ffdshow configuration? (they are disabled by default)

[Thunderbolt8]

whats that ffmpeg page again where I can also make feature requests?
^^ ???

thanks!

[Jeremy Duncan]

Please add this patch to the ffdshow svn: link (http://www.mediafire.com/?tmdwym5nnoo)

I got the patch from SEt: link (http://forum.doom9.org/showthread.php?t=148117)

I need a ffdshow generic build made that has this patch available.
The sooner the better. many thanks! :)

[leeperry]

Please add this patch to the ffdshow svn: link (http://www.mediafire.com/?tmdwym5nnoo)
hummm, that patch makes some of my movies run at half speed.

it'd be wiser to wait for Leak to watch over the problem(if there's any) whenever he has time I think....pushing dirty hacks into the SVN doesn't sound too wise :o

[Octo-puss]

Did you try enabling different adpcm choices in the ffdshow configuration? (they are disabled by default)

Yes I did, there is only libavcodec available anyway.

[lych_necross]

hummm, that patch makes some of my movies run at half speed.

it'd be wiser to wait for Leak to watch over the problem(if there's any) whenever he has time I think....pushing dirty hacks into the SVN doesn't sound too wise :o
I second that. AFAIK, Leak is already looking in to this issue.

[Blight]

Possible API Bug Report:
I believe that "IDFF_subCurrentFlnm" value determines whether ffdshow has found a suitable subtitle stream for rendering.
The problem is, this value is not set when an embedded sub stream is displayed, so there's no way of knowing whether ffdshow had found a valid subtitle stream and is ready to display it.
Is it possible to enable this flag when the embedded subtitle pin is connected?

[albain]

I fixed the channels order for Dolby TrueHD.

For LPCM, I will need Haruhiko's hand : channels order is wrong
Side left is sent to side right
Side left is sent to LFE
LFE is sent to side left.

Problem is located in TaudioUncompressed.cpp, line 100
if (lpcm24) {
if (highDef)
*dst++ = (src[i]<<24) + (src[i+ 1]<<16) + (src[i+ 2]<<8);
*dst++ = (src[i+3]<<24) + (src[i+ 4]<<16) + (src[i+ 5]<<8);
*dst++ = (src[i+6]<<24) + (src[i+ 7]<<16) + (src[i+ 8]<<8);
*dst++ = (src[i+9]<<24) + (src[i+10]<<16) + (src[i+11]<<8);
{

[Mercury_22]

I fixed the channels order for Dolby TrueHD.

For LPCM, I will need Haruhiko's hand : channels order is wrong
Side left is sent to side right
Side left is sent to LFE
LFE is sent to side left.

Problem is located in TaudioUncompressed.cpp, line 100
if (lpcm24) {
if (highDef)
*dst++ = (src[i]<<24) + (src[i+ 1]<<16) + (src[i+ 2]<<8);
*dst++ = (src[i+3]<<24) + (src[i+ 4]<<16) + (src[i+ 5]<<8);
*dst++ = (src[i+6]<<24) + (src[i+ 7]<<16) + (src[i+ 8]<<8);
*dst++ = (src[i+9]<<24) + (src[i+10]<<16) + (src[i+11]<<8);
{

Maybe I'm wrong but for LPCM I think you meant

LPCM channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to SIDE LEFT
2 - SIDE LEFT wrong mapped to LFE
3 - LFE wrong mapped to SIDE RIGHT

Test 7.1 LPCM (http://www.megaupload.com/?d=0HM2JCEP)
Anyway thanks for TrueHD I can't wait to test this :thanks:

[albain]

Yes that's correct

[Mercury_22]

It's my understanding that libdts has been upgrade / replace with libdca, (http://www.videolan.org/developers/libdca.html) am I wrong ? If I'm not shouldn't libdts be update ?

@albain Can you Please compile a x64 for testing TrueHD

[albain]

Possible API Bug Report:
I believe that "IDFF_subCurrentFlnm" value determines whether ffdshow has found a suitable subtitle stream for rendering.
The problem is, this value is not set when an embedded sub stream is displayed, so there's no way of knowing whether ffdshow had found a valid subtitle stream and is ready to display it.
Is it possible to enable this flag when the embedded subtitle pin is connected?

No, this value has been disabled a while ago

To get the current subtitles, instead of calling the sendmessage method with WPRM_GETPARAM and "IDFF_subCurrentFlnm",
call it with MSG_GET_CURRENT_SUBTITLES and 0

[MatMaul]

It's my understanding that libdts has been upgrade / replace with libdca, (http://www.videolan.org/developers/libdca.html) am I wrong ? If I'm not shouldn't libdts be update ?
libdts = libdca
and it has not been updated since 2006 or so.

[tetsuo55]

isn't libdca deader than a doornail?

I unberstood it the other way around, libdca got dropped in favor of libavcodec.

[clsid]

Not all software requires endless development. Certainly not small projects like libdca that have a fixed specification.

[leeperry]

libdts = libdca
and it has not been updated since 2006 or so.
yep, there's no DTS96/24 GPL decoder :(

[tetsuo55]

Not all software requires endless development. Certainly not small projects like libdca that have a fixed specification.True for a lot of applications, DTS has however evolved and libdca never catched up.

[clsid]

Those should be considered new formats, not plain old DTS. And those are also not yet supported by libavcodec or any other open source library.

[Mercury_22]

I fixed the channels order for Dolby TrueHD.

For LPCM, I will need Haruhiko's hand : channels order is wrong
Side left is sent to side right
Side left is sent to LFE
LFE is sent to side left.

Problem is located in TaudioUncompressed.cpp, line 100
if (lpcm24) {
if (highDef)
*dst++ = (src[i]<<24) + (src[i+ 1]<<16) + (src[i+ 2]<<8);
*dst++ = (src[i+3]<<24) + (src[i+ 4]<<16) + (src[i+ 5]<<8);
*dst++ = (src[i+6]<<24) + (src[i+ 7]<<16) + (src[i+ 8]<<8);
*dst++ = (src[i+9]<<24) + (src[i+10]<<16) + (src[i+11]<<8);
{

I've just tested your fix for TrueHD with clsid's ffdshow_rev3040_20090724_clsid and there is NO change :confused:
E-AC3 (Dolby Digital Plus) channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - SIDE LEFT wrong mapped to REAR LEFT

= No Side Channels

Test 7.1 E-AC3 (Dolby Digital Plus) (http://sharebee.com/69772daa)

Dolby TrueHD channels are wrong mapped :

1 - SIDE RIGHT wrong mapped to REAR RIGHT
2 - REAR RIGHT wrong mapped to SIDE RIGHT
3 - SIDE LEFT wrong mapped to REAR LEFT
4 - REAR LEFT wrong mapped to SIDE LEFT

Test 7.1 Dolby TrueHD (http://www.megaupload.com/?d=7P9P81CW)

[tetsuo55]

Those should be considered new formats, not plain old DTS. And those are also not yet supported by libavcodec or any other open source library.What i read about DTS states that all formats are plain DTS with extra data. This would mean that it's not a new format but an extention of a existing one. DTS even states that any version of DTS is fully backwards compatible with older players, the extra data of the newer format is simply discarded.

[clsid]

Logic dictates that it should decode DTS, not some extensions that were made up long after it was developed.

But it all does not matter. It does not support those extensions, neither do any other open source decoders. I can't change that. You can't change that. So we are all wasting our time here.