ffdshow tryouts project: Discussion & Development

[clsid]

The problem is caused by r4260 (http://ffdshow-tryout.svn.sourceforge.net/viewvc/ffdshow-tryout?view=revision&revision=4260), in which support for interlaced VC-1 was added.

[_pietro]

is it possible to use ffdshow with windows media player 12 to play dvds?!? i can't figure to work the mpeg2 option even disabling win7 codecs...thanks

[FlashGordon]

I think I've found a glitch with the MPEG2 libavcodec decoder. The skips happen when I open the DVD from the VIDEO_TS.IFO file. The glitch is 100% reproducible for me and the skips happen at the same exact spots. If I open the VOB by itself using LAV video decoder, the skips do not occur. If I open the VIDEO_TS.IFO using the MPC-HC mpeg2 decoder, the skips do not occur. Here is a sample: http://www.mediafire.com/download.php?im9htdjxo3v2kp7

EDIT: I tested the sample on my other machine and the skips are still there with latest ffdshow. Also, I realized depositfiles requires a 60 second wait, a captcha, and a 20 minute download time... so I updated the link to mediafire

[egur]

I've added HW deinterlacing, detail and denoise to the Intel QuickSync Decoder.
I don't want to commit my changes before there's some kind of GUI (in ffdshow) to control these.
My preferred option is to create a new tab instead of changing existing ones.
Can someone give me some pointers?

[clsid]

How may options are we talking about?
You could add a page below "Decoder options" if you need a lot of space. Or add to the "Decoder options" page if just a few options. The IDCT and quant options could be removed imo to free up some extra space on that page.

[egur]

I'd like to add a new page - below the decoder options is fine.
I want to add options for:
* HW DI on/off/force, half/full rate, etc.
* Denoise filter
* Detail filter
* procamp filter
These will only work with the Intel QuickSync Decoder - covers most of the video content these days.

[egur]

Committed several small changes:
Synced with latest QS decoder - video post processing is hard coded off until GUI is ready.
Fixed telecine manager - detected H264 frame doubling/tripling flags as 3:2.
Tweaked fps calculation - more stable.

[egur]

Last commit:
Added "Intel QuickSync" page below the decoder options.

[Delerue]

It seems that resize has broken DVD menus navigation again. :(

Edit: I mean, with 'Keep original aspect ratio' selected.

[wanezhiling]

http://forum.doom9.org/showpost.php?p=1574375&postcount=19549
Is there any chance to fix this?

[konok]

Hi,
Looking at the change list for build 4437, I see "remove vobsub support (except DVD sub)." listed.
What does this mean? Is backward compatibility with vobsub broken somehow?

I tried build 4453 and just noticed this change.

My problem... If I use just VobSub/vsfilter for IDX/SUB subtitles, I see double subtitles when playing SRT subtitles. If I disable ffdshow's text subtitles, it works, but I loose the nice extras in ffdshow. Does anyone have any ideas what I can do to make ffdshow the only one that displays for SRTs?

I've gone back to build 4422 for now.

[andybkma]

Hi, is it possible to set Profiles/Preset Settings by file extension (ie FLV, AVI, MP4 etc)? If so how to do you that as I don't see any option to select file extensions when I click "Preset Autoload Conditions..."

Cheers much :)

[jmartinr]

Hi, is it possible to set Profiles/Preset Settings by file extension (ie FLV, AVI, MP4 etc)? If so how to do you that as I don't see any option to select file extensions when I click "Preset Autoload Conditions..."
Try "on movie file name match (with wildcards)".

[Stanislav Mikhailenko]

Hello, used ffdshow_rev4422_20120409.exe

Accroding to spec for VC1 there is no restrict to change it during the stream http://rfc-ref.org/RFC-TEXTS/4425/chapter4.html
4.6. Repeating the Sequence Layer Header

Here is sample file that hangs on decoding the stream, using

WMVideo Decoder DMO it plays fine:
http://dl.dropbox.com/u/9992211/crash.wmv

The sequence layer and entry layer changes during the stream:


00 00 01 0f cb ee 13 f0 ef 88 80
00 00 01 0e 18 44 4f c3 bc 80
00 00 01 0f cb ef 13 f0 ef 0a 13 f8 3b f1 c0 ef 90 81 2d c6 db 8d 40
00 00 01 0e 4a 3f 28 40 00 00 01 0e 4a 3b 98 40


another one that gives broken image at the beginning,using WMVideo Decoder DMO it plays fine

http://dl.dropbox.com/u/9992211/broken.wmv

The sequence layer and entry layer changes during the stream:

00 00 01 0f ca 00 13 f0 ef 0a 13 f8 3b f1 c0 ef 88
00 00 01 0e 4c 10 80
00 00 01 0f cb ef 13 f0 ef 0a 13 f8 3b f1 c0 ef 90 81 2d c6 db 8d 40 00 00 01 0e 4a 3f 28 40
00 00 01 0e 4a 3b 98 40


The question: Is it problem of the decoder, that doesn't support changing of sequence layer, or in the stream itself?

[andybkma]

Try "on movie file name match (with wildcards)".

Cheers, got it sorted. *.flv, *.mp4 etc :-)

[Reino]

I believe it's been reported before, but anyway:
- http://media.nero.com/Standard720x576_1MBit.mp4 (http://media.nero.com/Standard720x576_1MBit.mp4)
AAC Audio part decoding failure with libavcodec. No problem though with libfaad2.Fixed.FFDShow has already abandoned libfaad2, but now (since rev3962 (http://mpc-hc.svn.sourceforge.net/viewvc/mpc-hc?view=revision&revision=3962)) MPC-HC has done so too.
Because, with the latest FFDShow and LAV Filters versions, the aac audio stream of this particular video is still incorrectly being decoded here, I was wondering if it could be a WinXP-only issue.Traps' movie trailer (http://forum.doom9.org/showthread.php?p=1578130#post1578130) plays fine for me too. But then I realized the issue I had with AAC Audio some time ago is gone now. Suddenly no problems anymore decoding Standard720x576_1MBit.mp4 with libavcodec. Don't know since when this is fixed and I have no idea what could've caused it, but anyway...good job! :)

[mandarinka]

Recently I noticed this post (http://forum.doom9.org/showpost.php?p=1541122&postcount=14179) implying that the software scaler in ffdshow doesn't work with 10-bit video. Is it a limitation of ffmpeg/libav's swscale, or is it just something that hasn't been enabled ffdshow-side?

[Px]

I want to ask, is it possible to add ability to prefer connection to ASIO Renderer (http://surodev.com/?p=451) to ffdshow audio filter (Output section, "Connect to:" menu)? It's CLSID is {8E14549A-DB61-4309-AFA1-3578E927E933}

[Blight]

Hi,
I've encountered an issue/bug when using ffdshow's subtitle renderer filter.
What I'm trying to do is build a graph where the ffdshow subtitle renderer is connected to the video decoder and the external subtitle file resides in a different directory entirely.

I'm trying to use the API to manually specify the subtitle file by calling the "IffDecoder2" interface like this:
putParam(IDFF_isSubtitles,Integer(True)); // Make sure subtitle processing is enabled.
putParam(IDFF_subAutoFlnm,Integer(False)); // Supposed to disable the auto-searching for sub files in the same dir as menu file.
putParamStrW(IDFF_subFilename,@SubFileLoad[1]); // Set the subtitle filename.

There are two issues:
1. The 'search in' radio-item remains checked rather than the 'file' radio-item becoming enabled and the subs are not rendered.
2. The IAMStreamSelect interface that lists the available subtitles doesn't see to be properly updated when using this method, even when manually switching to 'file' and clicking on the 'reload' button.

Am I doing something wrong? Can this be fixed?

[wanezhiling]

The project seems to be given up.. sigh..

[ryrynz]

The project seems to be given up.. sigh..

For the time being, Haruhiko is the only one that really works on it so it's dependent on the time he has available.

[obieobieobie]

Also it is summer. Many projects take breaks during the summer months.

[Budtz]

They've always been doing that. Who knows why.
Just get builds here:
http://xhmikosr.1f0.de/ffdshow/

Build with latest stable GCC (4.7.0)

How about the official tryouts site. are they also using old builds of the GCC? And arent the ones linked to above at least on paper less stable than the official build?

[clsid]

Official builds are build using the latest stable GCC (4.7.1) and MSVC 2010 SP1.

[naoan]

FFDShow fail to change preset if you try to change it during playback using only "OK" button. You need to change something so that apply button light up first, only then FFDShow will use the new preset.

[ENAJonas]

@clsid or Haruhiko - Do you know if this issue was fixed? I still have it on a very recent build, and I've up until now just been compensating for it by putting a static 250ms audio delay. It only happens on some 720p (or less) AVCs (but that's my primary media!!!) but I know for sure it is ffdshow video, and not any other issue. Thanks so much.

http://www.svp-team.com/forum/viewtopic.php?id=588

*Note - I'm not using SVP, just wanted the article. Issue is not SVP related.

[ENAJonas]

Friendly bump?

[ton4eff]

Hi, how to make ffdshow to display not only the current frame number, but total frame number on OSD? I'm using raw video filter. Like in this (http://imgbox.com/aarjPPmj) screenshot.

[fmagreed2]

Could support for the HDYC colorspace (input) be added to the encoder? It would be greatly appreciated.

HDYC is identical to UYVY except it uses the Rec. 709 colorspace instead of Rec. 601

Rec. 601 coefficients:
Y' = 0.2990 R' + 0.5864 G' + 0.1146 B'

Rec. 709 coefficients:
Y' = 0.2126 R' + 0.7152 G' + 0.0722 B'

Rec. 601:
Red x=0.670 y=0.330
Green x=0.210 y=0.710
Blue x=0.140 y=0.080
White x=0.310 y=0.316

Rec. 709:
Red x=0.640 y=0.330
Green x=0.300 y=0.600
Blue x=0.150 y=0.060
White x=0.3127 y=0.329

Hope I've provided enough information for someone to add this! Since it's so similar to UYVY, I'm hoping it's mostly a copy & paste job.

Any chance this could be added? I'd love to be able to use my black magic card with ffdshow.

[Joniii]

ffdshow DXVA surface overlay causes all the WMC controls, clock etc to flicker on Windows 8 Professional with Windows Media Center (RTM). Other than that everything works fine on W8. Build was "ffdshow_rev4475_20120712_clsid_x64".

[andybkma]

Greets, previously on older versions, there was a shortcut made to raw video filter settings (along with audio and video settings, so three settings shortcuts were made). How to get to the raw video filter settings now that that shortcut is not made anymore? Cheers

[ryrynz]

It was decided not to create shortcuts to raw anymore. Create a copy of the original ffdshow link and change the target to:
C:\Windows\SysWOW64\rundll32.exe "C:\Program Files (x86)\ffdshow\ffdshow.ax",configure ffdshowraw

[andybkma]

It was decided not to create shortcuts to raw anymore. Create a copy of the original ffdshow link and change the target to:
C:\Windows\SysWOW64\rundll32.exe "C:\Program Files (x86)\ffdshow\ffdshow.ax",configure ffdshowraw

Hi thanks for your help. I tried what you said, copied and pasted the address as the new target but didn't work. I only got the main ffdshow video settings, not the raw

Edit: Well I backdoored my way through ZoomPlayer so got it figured out that way. Still would be nice to figure out though why the shortcut didn't work

[ryrynz]

Apologies, my internet PC had the link wrong.

C:\Windows\System32\rundll32.exe "C:\Program Files (x86)\ffdshow\ffdshow.ax",configureRaw

[Reino]

As long as "Start in:" is pointing to the ffdshow directory,
%windir%\system32\rundll32.exe ffdshow.ax,configureRaw is enough for the Raw Video Decoder config
and
%windir%\system32\rundll32.exe ffdshow.ax,configureAudioRaw for the Raw Audio Decoder config.

[Vincent Vega]

guys how do i do that? in old ffdshow there was setting for that that worked, now it's gone on Output section. i have file with correct DAR at container level and incorrect at h264 stream level. i render in graphedit ffdshow first sends the correct DAR from container. then i start playback DAR changes to the incorrect that renderer (EVR custom) probably gets from stream. the above now gone setting used to help against this. so how do i fix it now?

[burfadel]

Latest Xhmikosr builds from http://xhmikosr.1f0.de/ffdshow/ don't work for me.

Every build before 4481 and earlier works fine, but the recent build 4484 and now the latest 4486 don't work at all. I use the ICL builds, haven't tried the others. Player freezes when loading the files, and the ffdshow plugin for avisynth crashes immediately if used. This is on a Core 2 Q9400 (has SSE 4.1 but not 4.2).

[nevcairiel]

Use the normal builds, ICL builds always had some issues, but nearly no advantages.

[burfadel]

Use the normal builds, ICL builds always had some issues, but nearly no advantages.

FFmpegsource build r706 just posted in this post:
http://forum.doom9.org/showpost.php?p=1588762&postcount=1612

Also doesn't work, I am guessing it' an issue relating to recent changes in Libav.

[andybkma]

Greets, got a coupla quick questions. Upgraded to a new laptop with Core i7-3610QM quad-core 2.3GHz Ivy Bridge CPU from my 5-year old Core-Duo (its my birthday hehe)

Under the ffdshow decoder options tab, there is an option to set "Number of decoding threads". What should I set this to for most efficient ffdshow use? 4? I had it set to 2 on my old laptop. Or it doesn't matter and ffdshow will automatically use the most efficient cores/process?

And second, I am now able to select this new thing called "Intel Quicksync" under H264 and VC-1. Ummm, I really didn't see a difference using it vs libavcodec/wmv9. Can you please let me know what benefits I am supposed to get from selecting Quicksync over the old "normal" ffdshow? Lower cpu use, better picture?

Note: I use mVR as my renderer. Cheers :-)

[egur]

Greets, got a coupla quick questions. Upgraded to a new laptop with Core i7-3610QM quad-core 2.3GHz Ivy Bridge CPU from my 5-year old Core-Duo (its my birthday hehe)

Under the ffdshow decoder options tab, there is an option to set "Number of decoding threads". What should I set this to for most efficient ffdshow use? 4? I had it set to 2 on my old laptop. Or it doesn't matter and ffdshow will automatically use the most efficient cores/process?

And second, I am now able to select this new thing called "Intel Quicksync" under H264 and VC-1. Ummm, I really didn't see a difference using it vs libavcodec/wmv9. Can you please let me know what benefits I am supposed to get from selecting Quicksync over the old "normal" ffdshow? Lower cpu use, better picture?

Note: I use mVR as my renderer. Cheers :-)

QuickSync will use less CPU on medium-high bitrate clips. Libavcodec will choke on very high bitrate clips. QS will allow you to play almost any H264 clips, up to 4K video.
Also works for MPEG2 and WMV9.

[clsid]

Latest Xhmikosr builds from http://xhmikosr.1f0.de/ffdshow/ don't work for me.

Every build before 4481 and earlier works fine, but the recent build 4484 and now the latest 4486 don't work at all. I use the ICL builds, haven't tried the others. Player freezes when loading the files, and the ffdshow plugin for avisynth crashes immediately if used. This is on a Core 2 Q9400 (has SSE 4.1 but not 4.2).The normal builds from SourceForge should work.

[mkanet]

Hi haruhiko_yamagata,

EDIT: Is there any plans to add native support for IVTC'ing both hard/soft telecine media files?

The reason I ask is because external AVISynth ITVC plugins such as TIVTC were not optimized for realtime IVTC processing.

Im still using Dscaler IVTC mpeg2 decoder; which is still the ONLY mpeg2 decoder in existence that can comfortably do this on most PCs. It would be nice if ffdshow could do this as well.

[Joniii]

ffdshow DXVA surface overlay causes all the WMC controls, clock etc to flicker on Windows 8 Professional with Windows Media Center (RTM). Other than that everything works fine on W8. Build was "ffdshow_rev4475_20120712_clsid_x64".

Forgot to mention that this doesn't happen with H.264, only with VC-1.

[kc7bfi]

We are using the Vega stream analyzer to analyze a H.264 stream. For almost ever AU we are seeing the error "forbidden_zero_bit value is incorrect in NAL. Found 1. Expected 0." Will this cause a problem with the ffdshow H.264 decoder when playing the video?
Thanks, David

[Joniii]

Im having a problem with Pirates of The Caribbean Blu-ray. Ripped mkv plays back vertically compressed with ffdshow DXVA. With default Windows decoder it is fine, also muxing the video to m2ts plays with correct AR with ffdsow DXVA. I'm not really sure if this is a problem with ffdshow or mkvtoolnix.

Here is a sample.
https://skydrive.live.com/redir?resid=5AA520AF9EAB7FD8!163

Also don't know if this is related to my problem:
http://www.makemkv.com/forum2/viewtopic.php?f=8&t=4264

[Liisachan]

I posted this report (https://sourceforge.net/apps/trac/mpc-hc/ticket/2588) to MPC-HC (including a sample MKV), but I think ffdshow's Subtitles filter shares the same problem.

When the font border is very small but not zero, the font shadow is not drawn, even if the shadow parameter is big, resulting in unintended shadow blinking when the border width is slowly animated from non-zero to 0, or from 0 to non-zero.

Though, I guess ffdshow might be intentionally bug-to-bug compatible with VSFilter...

[Reino]

@ Joniii:
With ffdshow (DXVA), MPC-HC reports 1920x1080 (AR 23:9) and the aspect ratio is incorrect. With ffdshow and MPCVideoDec, MPC-HC reports just 1920x1080 and everything's fine. Whether this is ffdshow (DXVA) to blame, or MakeMKV I don't know. But in the meantime, to get the right aspect ratio in MPC-HC, you can just select View -> Video Frame -> Override Aspect Ratio -> 16:9.

[kc7bfi]

We are using the Vega stream analyzer to analyze a H.264 stream. For almost ever AU we are seeing the error "forbidden_zero_bit value is incorrect in NAL. Found 1. Expected 0." Will this cause a problem with the ffdshow H.264 decoder when playing the video?
Thanks, David

From the h264.c file, the following references the forbidden bit:


const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src,
int *dst_length, int *consumed, int length)
{
int i, si, di;
uint8_t *dst;
int bufidx;

// src[0]&0x80; // forbidden bit
h->nal_ref_idc = src[0] >> 5;
h->nal_unit_type = src[0] & 0x1F;

src++;
length--;


If I am reading this correctly, if the forbidden bit is 1 not 0 then the nal_ref_idc value will be incorrect. Am I right? Will this cause problems further on down stream in the decoding processes?

David

[Guest]

If I am reading this correctly, if the forbidden bit is 1 not 0 then the nal_ref_idc value will be incorrect. Am I right? Yes.

Will this cause problems further on down stream in the decoding processes? It depends on whether nal_ref_idc is masked with 0x03 at future uses of it. As that seems unlikely (I haven't looked at the code), it's probable that this would cause issues with decoding.

Have you checked the streams to see if that bit is really set versus there being an issue with Vega stream analyzer? If you'd like to post a link to a sample, I could have a look at it.