ffdshow tryouts project: Discussion & Development

[haruhiko_yamagata]

I will send preset file now. I have the borders set to white so I can see what ffdshow is doing. In this case, there should be no white borders, but instead massive white borders are added at the top and bottom making the image squashed.
Thanks, I have reproduced.

[haruhiko_yamagata]

Thanks, here's another problematic sub (http://www.mediafire.com/download.php?zhmdmizlje2). It starts at 00:04 sec and subs will move incorrectly or overlap over each other.
\move(424,462,509,362,1057,1390)\t(1057,1390,\fscx80\fscy80)
\t is not implemented. It should get smaller while it is moving.
It is overlapped because the font is to big.

[haruhiko_yamagata]

Thanks a lot tal.aloni: now compilation is working (almost: only a problem with ff_vfw).
I'll look into how to use InnoSetup now for creating the install program.

In the end, I might not change/patch anything, but it will be my first compiled program!

Thanks haruhiko_yamagata... it's just a few minutes too late. :D
Looking at the wiki, I didn't download/install MinGW, but I had to download/install another program: yasm.exe.I have added yasm.
MinGW is required to compile libmplayer, libavcodec and some other dlls.

[haruhiko_yamagata]

Can you give me the access ?

Thanks
Please register yourself. Press Login and register.

[e-t172]

Just a small question to help me started: can I compile ffdshow with Visual C++ 2008 Express Edition, or do I need to find a full edition of Visual Studio 2008? Is this thread the best place to ask about how to compile ffdshow, or is there another place with help for that?

Using VC++ 2008 Express Edition I managed to build a ffdshow.ax which works flawlessly.

[clsid]

Do you know a way to contact Mike ? I have subscrbed to his mailing list but he rarely answers to it
Also it would be a good idea to implement such an option in gabest matroska splitter too ?

Thanks
Maybe BetaBoy knows how to contact him.

The Gabest splitters don't have a GUI with options, so adding a optional tweak would be more difficult. Also that Matroska splitter is inferior to Haali's so people should simply not use it (outside of MPC).

[haruhiko_yamagata]

Just back again to speak about aspect ratios. Definitely lots of problems relating to aspect ratio handling when resize is enabled. In my current config I can post a link to a file that shows the wrong aspect ratio.

http://www.dvdloc8.com/dolbydtsclip.php?clipid=3

This dtspiano.vob audio test which is 16:9 displays incorrect in the following setup.
Resize is not the problem. It's the stream that is wrong IMO. According to the decoders of ffdshow, the stream is 720x526, SAR 256/135, DAR 64/27.
The aspect ratio of MPEG-2 stream is quite unreliable according to FFmpeg. Currently there are two ways of calculating AR (see the code below).
If it is DVD, it is either 4:3 or 16:9. In that case, we can fix AR.
However, if it is a file, I don't know which way to choose.
Gurus of MPEG-2, please lead me to a correct fix.
//MPEG-2 aspect
if(s->aspect_ratio_info > 1){
//we ignore the spec here as reality does not match the spec, see for example
// res_change_ffmpeg_aspect.ts and sequence-display-aspect.mpg

/* ffdshow custom code - we do not want to ignore spec. Comment out "// 1".
DVD plays better. It breaks the playback of res_change_ffmpeg_aspect.ts though. */
if( (s1->pan_scan.width == 0 )||(s1->pan_scan.height == 0) /*|| 1*/){
AVRational r={s->width, s->height};
s->avctx->sample_aspect_ratio=
av_div_q(
ff_mpeg2_aspect[s->aspect_ratio_info],
r//(AVRational){s->width, s->height}
);
}else{
AVRational r={s1->pan_scan.width, s1->pan_scan.height};
s->avctx->sample_aspect_ratio=
av_div_q(
ff_mpeg2_aspect[s->aspect_ratio_info],
r//(AVRational){s1->pan_scan.width, s1->pan_scan.height}
);
}

[haruhiko_yamagata]

Using VC++ 2008 Express Edition I managed to build a ffdshow.ax which works flawlessly.
Great, I have updated the wiki.

[albain]

Maybe BetaBoy knows how to contact him.

The Gabest splitters don't have a GUI with options, so adding a optional tweak would be more difficult. Also that Matroska splitter is inferior to Haali's so people should simply not use it (outside of MPC).

Okay thanks, I have PM him.

[mark0077]

Resize is not the problem. It's the stream that is wrong IMO. According to the decoders of ffdshow, the stream is 720x526, SAR 256/135, DAR 64/27.
The aspect ratio of MPEG-2 stream is quite unreliable according to FFmpeg. Currently there are two ways of calculating AR (see the code below).
If it is DVD, it is either 4:3 or 16:9. In that case, we can fix AR.
However, if it is a file, I don't know which way to choose.
Gurus of MPEG-2, please lead me to a correct fix.

Hi thanks for looking into the issue. If it is the stream that is wrong, what about the DVD's like the Test DVD. It appears in the perfect aspect ratio when not using ffdshow resize. When resize is used, it is squashed horizontally.

[haruhiko_yamagata]

Hi thanks for looking into the issue. If it is the stream that is wrong, what about the DVD's like the Test DVD. It appears in the perfect aspect ratio when not using ffdshow resize. When resize is used, it is squashed horizontally.
In DVD applications, video renderer correct AR if resize is not used. Please use ffdshow beta6 + OSD (User:%outd %odim %oar).
If AR is wrong, menu does not work though.

[mark0077]

In DVD applications, video renderer correct AR if resize is not used. Please use ffdshow beta6 + OSD (User:%outd %odim %oar).
If AR is wrong, menu does not work though.

Do you need me to gather more info with beta 6 using DVD's that show the wrong aspect ratio? I am concerned because all display correctly without resize, many display incorrectly with resize.

[haruhiko_yamagata]

Do you need me to gather more info with beta 6 using DVD's that show the wrong aspect ratio? I am concerned because all display correctly without resize, many display incorrectly with resize.
I'm just explaining that DVD applications can show wrong AR correctly.

[BetaBoy]

Albain... thx for the detective work on the filter issue in WMP/MCE and Windows 7. I started a dedicated thread here on D9 ( http://forum.doom9.org/showthread.php?p=1265646 ) on what your initial findings were, so others could confirm if this is the case. If so... well... I'm speechless.

[haruhiko_yamagata]

Is video_format in MPEG-2 reliable?
video_format specifies component/PAL/NTSC/SECAM/MAC/...
If it is reliable,
if (video_format== PAL && width==768 && height==576){
// Force 4:3 or 16:9
}
if (video_format== NTSC && width==720 && height==480){
// Force 4:3 or 16:9
}is possible.

[mark0077]

Is video_format in MPEG-2 reliable?
video_format specifies component/PAL/NTSC/SECAM/MAC/...
If it is reliable,
if (video_format== PAL && width==768 && height==576){
// Force 4:3 or 16:9
}
if (video_format== NTSC && width==720 && height==480){
// Force 4:3 or 16:9
}is possible.

I suppose if it is possible to check what the renderer does to decide then you could use that as a starting point. I think I understand the problem now.

[Amour]

MinGW is required to compile libmplayer, libavcodec and some other dlls.

Ok, true, I was slow to notice some important missing files. I'll study how to compile them as well.

By the way, on the wiki, you write "Microsoft Platform SDK" (aka Microsoft Windows SDK). But are you sure it is required? I haven't installed it for building the vcproj.

[clsid]

The Platform SDK is not needed. Visual Studio comes with its own SDK.

[fastplayer]

albain, you've made it into the news section of heise.de! Congrats! :)
Heise is the German publisher of the very reputable PC magazine "c't".
Here's the link: Windows 7 blockiert Codecs von Drittanbietern (http://www.heise.de/newsticker/Windows-7-blockiert-Codecs-von-Drittanbietern--/meldung/135209)
I'l post a link to the English translation, once it gets available.

[albain]

Good news :)

The most buzz we make, easier it will be to make Microsoft change his mind

[STaRGaZeR]

Is there a faster way of compiling all the stuff that requires GCC (libavcodec, ffmpeg-mt, etc.) that manually going to the required folder in MSYS and running the makefile?

[Thunderbolt8]

I know why I stick to XP. vista and 7 just seem to be useless lulz.

[clsid]

Is there a faster way of compiling all the stuff that requires GCC (libavcodec, ffmpeg-mt, etc.) that manually going to the required folder in MSYS and running the makefile?
You don't need MSYS. You can also run "make" from the windows command prompt if you use my MinGW32 installer.

[STaRGaZeR]

Cool, it works just fine. BTW it's impossible to make ICL builds with VS2008 Express Edition if anyone is interested.

[Amour]

Good to know, Stargazer. :)

I haven't installed MinGW32 yet, and I am first trying with VS2008 to compile the projects. Unfortunatly, for the following file I get multiple errors:
\trunk\src\codecs\x264\common\macroblock.c

Here are the first errors (there are a few more similar):
.\common\macroblock.c(842) : error C2143: syntax error : missing ';' before 'type'
.\common\macroblock.c(843) : error C2143: syntax error : missing ';' before 'type'

And the related lines:
int buf_hpel = (h->param.i_width+48) * sizeof(int16_t);
int buf_ssim = h->param.analyse.b_ssim * 8 * (h->param.i_width/4+3) * sizeof(int);

[Leak]

I haven't installed MinGW32 yet, and I am first trying with VS2008 to compile the projects. Unfortunatly, for the following file I get multiple errors:
\trunk\src\codecs\x264\common\macroblock.c
I'm pretty sure that's because nobody has been using the ff_x264 solution that's been last touched about 900 revisions ago for a long time.

Compiling it with GCC works fine...

np: SCSI-9 - Vesna, Lastic & Elliott (Easy As Down)

[haruhiko_yamagata]

You don't need MSYS. You can also run "make" from the windows command prompt if you use my MinGW32 installer.
I use MSYS. Without MSYS, what do you use as an alternative of 'rm'.

[mark0077]

Hi, Just playing around with building ffdshow also. I can compile fine now but where can I find the command to create the installer so I can test any changes I may make.

Also thanks haruhiko for fixing 95% of my dvd menu problems. Only some slight delays now, mostly when using external mpeg2 decoder. But its still perfectly usable now.

[jmartinr]

Subtitle Problem.

I'm sorry to say that "horizontal position 7%" with "alignment left" doesn't work as expected in revision 2815. The alignment does not make any difference now. It did work in revision 2737.

[Leak]

Hi, Just playing around with building ffdshow also. I can compile fine now but where can I find the command to create the installer so I can test any changes I may make.

Also thanks haruhiko for fixing 95% of my dvd menu problems. Only some slight delays now, mostly when using external mpeg2 decoder. But its still perfectly usable now.
Install InnoSetup then open the ffdshow_installer.iss file in bin\distrib\InnoSetup with it, maybe update the version and date at the top and hit "Compile" and "Run"...

np: Sylvain Chauveau - Nuage III (Pop Ambient 2009)

[STaRGaZeR]

Good to know, Stargazer. :)

I haven't installed MinGW32 yet, and I am first trying with VS2008 to compile the projects. Unfortunatly, for the following file I get multiple errors:
\trunk\src\codecs\x264\common\macroblock.c

Here are the first errors (there are a few more similar):
.\common\macroblock.c(842) : error C2143: syntax error : missing ';' before 'type'
.\common\macroblock.c(843) : error C2143: syntax error : missing ';' before 'type'

And the related lines:
int buf_hpel = (h->param.i_width+48) * sizeof(int16_t);
int buf_ssim = h->param.analyse.b_ssim * 8 * (h->param.i_width/4+3) * sizeof(int);

Forget about VS2008 for the GCC stuff, just install clsid's MinGW32 installer and from the command prompt go to each project folder and run "make". You can cook a simple .cmd and compile everything by just double-clicking it. Using InnoSetup is trivial.

BTW developers, I've noticed that my build of rev2812 has different file sizes for almost all .dll's compared to clsid's, some are bigger and some are smaller, while working just fine. Any reason for this?

[haruhiko_yamagata]

Subtitle Problem.

I'm sorry to say that "horizontal position 7%" with "alignment left" doesn't work as expected in revision 2815. The alignment does not make any difference now. It did work in revision 2737.
I have fixed at rev 2816.

[jmartinr]

I have fixed at rev 2816.

:thanks:

[clsid]

I use MSYS. Without MSYS, what do you use as an alternative of 'rm'.
The "UnxUtils" item in my MinGW pack installs rm.exe to the bin dir, which can be called from the command prompt :)

BTW developers, I've noticed that my build of rev2812 has different file sizes for almost all .dll's compared to clsid's, some are bigger and some are smaller, while working just fine. Any reason for this?I use ICL for compiling a couple of files (libmad, libfaad, libdts, etc) so that should explain the difference in size.

[haruhiko_yamagata]

BTW developers, I've noticed that my build of rev2812 has different file sizes for almost all .dll's compared to clsid's, some are bigger and some are smaller, while working just fine. Any reason for this?
You don't have to worry about it as long as it is working fine. It depends on the version of libraries, hearer files and compilers. In this complicated project, it is hard to match all of them completely.

[mark0077]

BUG: I notice recently that my keyboard shortcut in ffdshow CTRL+ALT+D for enabling / disabling de-interlacing, now causes ffdshow and mpc-hc to crash. Crash doesn't happen when manually clicking de-interlacing on or off so I imagine the issue is with the shortcut somehow.

[STaRGaZeR]

OK, thanks. rev2812 ICL11 build: http://www.zshare.net/download/57712418624f949b/

After rev2812 compiling with ICL is not possible because it complaints about operand types being incompatible in files coming from the subtitle branch:

http://thumbnails12.imagebam.com/3088/6ee08d30874987.gif (http://www.imagebam.com/image/6ee08d30874987)

[haruhiko_yamagata]

BUG: I notice recently that my keyboard shortcut in ffdshow CTRL+ALT+D for enabling / disabling de-interlacing, now causes ffdshow and mpc-hc to crash. Crash doesn't happen when manually clicking de-interlacing on or off so I imagine the issue is with the shortcut somehow.
OK, it's a bug in short OSD. I'll fix this.

[haruhiko_yamagata]

OK, thanks. rev2812 ICL11 build: http://www.zshare.net/download/57712418624f949b/

After rev2812 compiling with ICL is not possible because it complaints about operand types being incompatible in files coming from the subtitle branch:

http://thumbnails12.imagebam.com/3088/6ee08d30874987.gif (http://www.imagebam.com/image/6ee08d30874987)

Please try this (http://www.nabble.com/BOOST_FOREACH-not-compiling-with-Multindex-containers,-Intel-Compiler-version-11-td20541659.html).
It's up to boost or Intel.

[STaRGaZeR]

Please try this (http://www.nabble.com/BOOST_FOREACH-not-compiling-with-Multindex-containers,-Intel-Compiler-version-11-td20541659.html).
It's up to boost or Intel.

Yes, ICL10.1 doesn't have the problem. I will use it for the time being. Thanks.

[Amour]

haruhiko_yamagata, could you allow registered users to create new pages on the wiki? Thanks.

[ikarad]

I would like to know if the bugs of ffdshow-Mt will be soon corrected or not because in my blu-ray there is lot of bugs?

[haruhiko_yamagata]

haruhiko_yamagata, could you allow registered users to create new pages on the wiki? Thanks.
OK, updated permission.

[fastplayer]

haruhiko_yamagata, could you allow registered users to create new pages on the wiki? Thanks.
Wiki is down for "maintenance". Try again in half an hour.

[mark0077]

Just did a re-install of vista 64 lastnight and was playing some 1080p vc-1 files. Installed ffdshow and decided to try its vc-1 libavcodec decoder again to see if I was getting the terrible frame rate.

Yes I still get the problem with build 2815. Its MUCH better when "Drop frame on delay" is disabled as before. I get no frame drops recorded in mpc-hc. I am only getting 18% cpu usage, almost identical to the mpc-hc decoder. The only difference is terrible performance, looks like 10fps instead of 24.

Anyone know what might be wrong with both
vc-1 terrible performance compared to mpc-hc decoder even though same cpu usage.
"Drop Frame On Delay" causing more slowness even though no frame drops recorded.

[clsid]

I would like to know if the bugs of ffdshow-Mt will be soon corrected or not because in my blu-ray there is lot of bugs?You would have to ask Mr VacBob, he is the developer of ffmpeg-mt.


@mark0077
Do you get the same problem if you play those files in GraphStudio or GraphEdit? Just to rule out MPC related issues.

[mark0077]

You would have to ask Mr VacBob, he is the developer of ffmpeg-mt.


@mark0077
Do you get the same problem if you play those files in GraphStudio or GraphEdit? Just to rule out MPC related issues.

Neither of those programs can connect to my remote graph, no graph appears.... I have tried mpc-hc beliyaals build and normal one, using ever and evr-cp.... Strange thing is the lack of reporting of dropped frames. I would say over half the frames are dropped but not being recorded.

EDIT: OK I tested with reclock on and off, no difference BUT I notice something in reclock that may indicate where the problem is. Decoding in mpc-hc, reclock says it finds the frame rate 23.97fps. But decoding using libavcodec or wmv9 decoders in ffdshow, reclock doesn't know the frame rate..... This may be something to do with the problem.... I don't know what though, do you think this is ffdshow or mpc-hc related?

[clsid]

I meant letting GraphStudio do the graph building and playback. So no use of MPC at all. Then you can also manually alter the filter chain, for example to remove reclock.

[mark0077]

I meant letting GraphStudio do the graph building and playback. So no use of MPC at all. Then you can also manually alter the filter chain, for example to remove reclock.

This might take a while, been ages since I played with graph editing, not sure what splitter to split a .ts file..... Can any of you test vc-1 internal decoders with mpc-hc?

[ikarad]

You would have to ask Mr VacBob, he is the developer of ffmpeg-mt.


.

Do you have e-mail of Mr VacBob?