Is there any directshow filter for realmedia?

[zedude]

gabest, you rule :) thank you so much !!

it works incredibly much better now and i've made this graph (http://perso.wanadoo.fr/zedudeishere/site_images/matroska_graph2.jpg) ( it's something i used to dream of ^^ )
and it plays like a charm :D

[Assault]

Do I have to deactivate anything in Media Player Classic for using RealMediaSplitter? Because on my system MPC doesn't use RealMediaSplitter for rmvb-file playback but it still uses the "old" method as it could always playback RealMedia files. Furthermore I can't playback Gabest's matroska file in MPC but it is working in other direct show players like Zoomplayer or Windows Media Player. Does anyone have similar problems?

Assault

[S_O]

S_O, what version of windows are you using? could give more specs on you computer. that could help to solve.
Windows 2000 SP3
Athlon C 1333MHz
512MB DDR-Ram
Epox 8K7A (AMD Northbridge, VIA Southbridge)
Elsa Gladiac MX (Geforce2 MX)
Terratec DMX Xfire 1024
DirectX 9.0a

I also tried the new version, without any success. I´m able to play rm with dnet-audio, works perfectly (except this quality-bug in ac3filter), but no other audio-codec works and video doesn´t work at all. Also .ra (RealAudio) files doesn´t work. RealOne v2 is installed.

[wata]

i try the mkv file with latest RealMediaSplitter, it only play audio but not the video
(my video card is matrox g400 dh , OS is win2000sp3)

pop up this

Stream 1
RealVideo Decoder
AVI Decompressor
Color Space Converter
VGA 16 Color Ditherer

Media Type 0:
--------------------------
AM_MEDIA_TYPE:
majortype: MEDIATYPE_Video {73646976-0000-0010-8000-00AA00389B71}
subtype: MEDIASUBTYPE_RGB4 {E436EB79-524F-11CE-9F53-0020AF0BA770}
formattype: FORMAT_VideoInfo {05589F80-C356-11CE-BF01-00AA0055595A}
bFixedSizeSamples: 1
bTemporalCompression: 0
lSampleSize: 153600
cbFormat: 1128

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

BITMAPINFOHEADER:
biSize: 40
biWidth: 640
biHeight: 480
biPlanes: 3
biBitCount: 4
biCompression: 2000928
biSizeImage: 153600
biXPelsPerMeter: 0
biYPelsPerMeter: 0
biYPelsPerMeter: 0
biClrUsed: 16
biClrImportant: 16

[Blight]

Great work gabest.

And let it not be said that great work goes unrewarded, check PM.

[iwod]

Originally posted by Blight
Great work gabest.

And let it not be said that great work goes unrewarded, check PM.

and i would like to know what is in that pm........ :D :D

[unmei]

some nice errors i collected (on XP):

bs player
-latest matroskaMuxer not installed, real one (.dlls) not installed -> crash
-real one not installed, but Muxer installed -> "unknown format" message, no play

MPC
-real one (.dlls) not installed -> only audio (ogg) plays
-"options->filters->source->matroska splitter" not unchecked -> only audio (ogg) plays

TCMP
-real one (.dlls) not installed -> only audio (ogg) plays

zoom player
-real one (.dlls) not installed -> only audio (ogg) plays

BTW with now both installed and internal matroska splitter unchecked in MPC it works great. Thanks gabest ! the major reason for me not to use rv for low bitrates (limited to ra and player) are gone at once :D

too sad the post processing is not adjustable in rv decoder but i'm confident some nerd will even figure this out someday

AFAIK candy cannot yet be sent with PM

[Ghim]

The only drawback I still see in this filter is that we got to install realone first so that real dll will be present but we can't do anything about it (It would be great if real did a dll only version but I doubt they'll do it...)

It's really a great filter you did here...
I'll be able to use any player to read Real Media Files and benefits of their advantages...

[Schultz]

Originally posted by Assault
Do I have to deactivate anything in Media Player Classic for using RealMediaSplitter? Because on my system MPC doesn't use RealMediaSplitter for rmvb-file playback but it still uses the "old" method as it could always playback RealMedia files. Furthermore I can't playback Gabest's matroska file in MPC but it is working in other direct show players like Zoomplayer or Windows Media Player. Does anyone have similar problems?

Assault

Make sure when you are Using MPC that you go into the Options and go to the Filters Section and Uncheck the Mastroska Splitter because the one that is built into MPC in the lastest release cannot play real media files inside mkv's .. But use the one in the filters file that gabest posted.

[midiguy]

hey so.. I know I asked before, but will this directshow filter solve realmedia jitter problems?

[haibane]

I have some playback problem with rm muxed into matroska, when the cpu usage reached 100%, the video will freeze, it happens when dragging window or exit fullscreen. I tried with the blade 2 clip and a clip I made. Another problem is that rmvb made with the dup filter have a/v sync problem. I test this with the same clip, one made with dup, one w/o.......
I have windows XP, and I'm using the filters in the previous page.........

[Animaniac]

Using a similar method, could QuickTime formats, such as Sorenson video, be placed in Matroska and use the QuickTime DLLs (or a future version of ffdshow) for playback?

[gabest]

kilg0r3: I don't know either how good they are, but they can be muxed into mkv. And there was that 6 channel sample posted by karl which is already supported.

Assault: mpc will open it the "old" way if it can recognize it as realmedia based on the file extension. The prob with mkv was answered by Schultz.

S_O: Did you tested it in graphedit with the overlay mixer? .ra will be supported later, it looks to be a little different file format, but I only peeked it once.

wata: Just what I told to S_O, try with the overlay mixer under all pre-xp windows versions.

unmei: Yea, updated matroska splitter and the codecs from realone are needed, this isn't really surprising :) About real's post processing I don't know too much... It might be either realone doing this after receiving the same raw decompressed picture as the dshow filter, or the codec can be controlled and I use it in some default mode.

midiguy: There are more renderers to choose from in dshow, you may find better or worse implementations than what you can in realone.

Animaniac: I still have a qt video decompressor filter somewhere, but it connected to the old mov splitter of dshow and that couldn't handle newer formats. I need to get specs on the file format first and rewrite that. But 3ivx also has something for quicktime, which I haven't tested yet.

haibane: Do you have a sample with a/v desync for me? It can be very low resolution to keep the file size small. For the other problem, isn't the picture freezing because the cpu can't decompress it in real-time?

[karl_lillevold]

Wow, all these experiments with new features suddenly now possible with Gabest's RealMediaSplitter! Looks like I have some testing and reading up on matroska, to do myself :)

Another interesting new feature is that Producer can now transcode RealMedia files. Previously this was not possible, but since Producer uses DirectShow to read its source files, via Gabest's filter, it appears to transcode just fine [with the exception of one sample encoded with DropDupe, don't know why]. Not well tested though, and transcoding from an already compressed source is generally not recommended either. Still neat.

All my test files play back very well with RealMediaSplitter, up to and including HD 1280x720p at 5 Mbps (overlay mixer required instead of system default renderer). There is still a problem with anamorphic files, like Gabest mentioned. Now I am looking forward to not having to change the (rm/rmvb) extension, such that MPC opens the files using this filter instead of the old method. Maybe an option will be added to MPC to select internal or external method.

[gabest]

Originally posted by karl_lillevold
Now I am looking forward to not having to change the (rm/rmvb) extension, such that MPC opens the files using this filter instead of the old method. Maybe an option will be added to MPC to select internal or external method. Yes, there will be a new column in the format list control to select the opener api (dshow, real, qt, swf).

[bond]

@karl
do you know if post-processing is done by the player or by the codec?

[Assault]

@ Gabest, Schultz

Thanks for clearing that up. It works now. ;)
Great work Gabest!!!

Assault

[shitowax]

Originally posted by gabest
Animaniac: I still have a qt video decompressor filter somewhere, but it connected to the old mov splitter of dshow and that couldn't handle newer formats. I need to get specs on the file format first and rewrite that. But 3ivx also has something for quicktime, which I haven't tested yet.
[/B]

Hi Gabest,
The 3ivx D4 404 splitter has some sync issues, contact me if you want to play with the new beta-one. Another thing, it would be a bit a waste of time to rewrite a qt format support from scratch as there is already at least 7 open source implementations of the mov format: the 3ivx splitter uses OpenQuicktime with is funky overloading mechanism, but there are tons of others implementations (QT4L, LibQT, ffmpeg, xine, mplayer, VLC have all their own libs)...

[ChristianHJW]

Originally posted by shitowax Another thing, it would be a bit a waste of time to rewrite a qt format support from scratch as there is already at least 7 open source implementations of the mov format

Yes, but if Gabest feels he needs a little 'finger training' he is rewriting a complete library from scratch in only 3 days ;) .... as happened for matroska :eek: :scared: ....

[Blight]

Maybe someone in the Helix team can point Gabest ot the Anamorphic Flag in the file format so that he can set the VIDEOINFOHEADER2 information to include the aspect ratio adjustments required for anamorphic content.

I tested the 3ivx codec a bit, it seemed to work pretty well with the limited content I have here, so repeating the work on the splitter is probably pointless. The bigger question is wether to use the QuickTime SDK to decode the content, or go through ffmpeg? sdk which I believe already supports sorenson.

[gabest]

Just added a little bilinear resizer to restore the display size, it is not very fast, but should be ok. Also enabled the video decoder to connect to the old renderer directly, but if your video card is broken and can't let it dynamically switch to yuy2 then only blackness will fill out the video window.

http://prdownloads.sourceforge.net/guliverkli/RealMediaSplitter_20030710.zip?download

[Dark-Cracker]

Hi

for anamorphic why not simply get the height and width from the .rmvb file and report them in the header of the matroska file (and use the anamorphic plugin) ? in anamorphic realmedia only the width or height are used to strech the video.

Does a tool was created to connect dshow pin ?


Ps: i have some slowdown during payback your balde2 sample is it normal ?

Bye.

[gabest]

Originally posted by Dark-Cracker
Hi

for anamorphic why not simply get the height and width from the .rmvb file and report them in the header of the matroska file (and use the anamorphic plugin) ? in anamorphic realmedia only the width or height are used to strech the video.I have no idea how to get the size, and because the codec returns it, it might be different for every frame, though very unlikely :)Does a tool was created to connect dshow pin ?This question I don't understand.Ps: i have some slowdown during payback your balde2 sample is it normal ? If the cpu is at 100% then yes, if not then no :D Which sample did you refer to? mkv, rm or both?

[S_O]

S_O: Did you tested it in graphedit with the overlay mixer? .ra will be supported later, it looks to be a little different file format, but I only peeked it once.
I tested the video with GraphEdit: plays fine. Manually inserted "Overlay Mixer" in front of the Video Renderer: plays fine. "Overlay Mixer 2": also fine. Tested it with WMP7 (yes, this crappy m$-player): perfect. Why doesn´t it work with wmp6.4?

[bond]

plays fine on my system (no more "no video" problems)!

but i get (slightly) choppy video playback (didnt had that with the first version of your filter)
perhaps that's the same like dark-crackers slowdown -> i get a real high cpu usage

i used the rm sample

[Dark-Cracker]

I have no idea how to get the size, and because the codec returns it, it might be different for every frame, though very unlikely.


i perhaps misunderstand your problem with anamorphic file, i have understand this (with an imaginary exemple) : u have an .rmvb file u encode it at real size : 640*400 and u use anamorphic to strech the file at 640*480 the size in the header is 640*480 but i suppose (i am perhaps wrong) the problem with your filter is when u convert the .rmvb in .mkv u get the raw frame resolution (640*400) i simply suggest u to get the size in the header and to calc anamorphic ratio (640/480) and change it in the output .mkv . but perhaps i have misunderstand your problem with anamorphic input.

This question I don't understand.

i simply ask if u think made a tool (like the koepi oggmux tool) to connect your dshow filter, or this will be added in mkvmerge tool.

If the cpu is at 100% then yes, if not then no Which sample did you refer to? mkv, rm or both?

no problems with .rm file but with mkv sample i have a slowdown every approximately 5 seconds (perhaps this filter need more cpu power).

PS: a little suggestion perhaps should u put in your first post all the interesting link to avoid to scroll each pages.

Ps2: sorry for my poor english.

Bye.

[shitowax]

Originally posted by Blight
I tested the 3ivx codec a bit, it seemed to work pretty well with the limited content I have here, so repeating the work on the splitter is probably pointless. The bigger question is wether to use the QuickTime SDK to decode the content, or go through ffmpeg? sdk which I believe already supports sorenson.

As far as I know ffmpeg supports Sorenson Video 1 and 3, but doesn't support QDesign Audio that is used very often in Quicktime trailer (even if some are already using AAC)... I think that a directshow wrapper for qt codecs can't be really avoided if you want the full compatibility ...

Originally posted by ChristianHJW
Yes, but if Gabest feels he needs a little 'finger training' he is rewriting a complete library from scratch in only 3 days .... as happened for matroska .... [/B]

I would say that the mov format isn't the real problem (it might be similar in complexity to matroska). The real problems is the private data inserted by codecs in the file header itself and required to decode... some are publicly available (MPEG-4 audio, MPEG-4 video), some are not : Sorenson, Qdesign, Pixlet, ... Another good one is all the small variants existing between the pure mov file format and all its derivative (mp4, m4a, 3gp, ...)

anyway, as usual, just my two cents.

[gabest]

S_O: Did you test the new release from today?

bond: I almost went to take a nap, but something suddenly came into my mind. The last two versions will start dropping frames when the rendering is late by at least half a second. While the playback with first release could have looked more smooth it wasn't actually prepared for real-time. If there had been sound too, the picture would have become desynced at the point where the new decoder makes it choppy to keep up with the sound.

Dark-Cracker: After such a file is transmuxed, matroska's file header will also say 640x480 because the rm splitter can only pass that to the matroska muxer. The real size will be only known by the decoder after it received back the first decoded frame from the codec dll.

[ChristianHJW]

Originally posted by gabest Dark-Cracker: After such a file is transmuxed, matroska's file header will also say 640x480 because the rm splitter can only pass that to the matroska muxer. The real size will be only known by the decoder after it received back the first decoded frame from the codec dll.

Similar problem as with MPEG4 in AVI. If the container doesnt know about the output AR its hard to support this on DirectShow, as the output window size has to be known already when building the graph.

The only possible workaround is add a resizing filter to the decoder filter itself, resize the actual window down to achieve the actual DAR, and add black borders on top and bottom of the active window to fill out the originally called output window size.

Of course, this is completely defeating the purpose of using anamorphic encoding, as even in full screen mode the video will first be resized to a smaller size, and thus all the extra vertical resolution gets lost, before being expanded to full screen from DirectShow.

Gabest, i recommend to use a very sharp resizing filter like Lancos at the very least, instead of bilinear.

wait !! .... me has an idea :D !!!

Why not call the actual output window from the DirectShow graph always 50% bigger than necessary ???

Say, you have a video with 640 x 400, and you dont know if its encoded anamorphic or not, only the codec will know as it's the only one to actually read the DAR flag from the Real stream. Why not call the graph to a 960 x 600 output window in any case on playback, so if you find the DAR flag and the decoder filter knows about the actual output size, you dont have to squeeze down the vertical resolution ?

Most modern monitors will work on a 1024 * xxx or 1280 * xxs res, so they should be fine in any case ? Can you make this configurable in the dshow decoder filter setup even ? What you think ?

[gabest]

The only possible workaround is add a resizing filter to the decoder filter itself, resize the actual window down to achieve the actual DAR, and add black borders on top and bottom of the active window to fill out the originally called output window size.The decoder can't and doesn't do this. It is only scaling up from the encoded size to the display size, which is the one taken from the header and used in the connection media type. From the display size it is the renderer and directx doing the stretching to the actual window size.

[bond]

Originally posted by gabest
bond: I almost went to take a nap, but something suddenly came into my mind. The last two versions will start dropping frames when the rendering is late by at least half a second. While the playback with first release could have looked more smooth it wasn't actually prepared for real-time. If there had been sound too, the picture would have become desynced at the point where the new decoder makes it choppy to keep up with the sound.hm interesting!
but any chance to get the cpu usage down to a lower level so that late rendering/frame dropping doesnt occur :(

[gabest]

With the overlay mixer and without any resizing in the filter, it should be as fast as realone. Is realone still playing the same clip smoother at you?

[ChristianHJW]

Originally posted by gabest The decoder can't and don't do this. It is only scaling up from the encoded size to the display size

:confused: :confused: ???

So, does that mean it is possible that the decoder filter will output a different resolution than the player told DirectShow when building the graph semiautomatically ?

If this is the case, what is the actual problem milan mentioned when talking about supporting the MPEG4 DAR from ffdshow ? It has a resizing filter built in and will read the DAR from the MPEG4 ES headers as soon as the first frame arrives, so whats the problem ??

BTW : bilinear sucks for scaling UP, its mainly suited for downsizing, you really should consider bicubic or lanzcos ;) ...

[karl_lillevold]

Originally posted by bond
@karl: do you know if post-processing is done by the player or by the codec?
This has been discussed a few times. All image processing in RV9 is inloop, thus non-separable and non-adjustable. The best known example of another codec having a similar feature, is H.264. The main advantage is that the codec gets to work with as clean frames as possible, increasing compression efficiency and video quality. The disadvantage is that it can not be adjusted without changing the bitstream, requiring the adjustment to be made by the encoder. This said, maybe one day for experimental purposes, I may make an adjustable encoder version, requiring a decoder update to go along with it.

[S_O]

S_O: Did you test the new release from today?
Installed the newest build, and: Perfect! It works great!
If Milan would be as fast as you and finally make a new version with rv10 and 14_4/28_8 support (which is in ffmpeg) there would be nearly complete RealMedia-support!
Nearly: sipro voice (sipr) is not played. Is there any way to decode it with "ACELP.net Sipro Lab Audio Decoder"-filter? I tried to mux the sipro audio into a wav and change identification to 0x130, but it doesn´t work, because the remuxed wavs doesn´t contain a 44 bytes-standard wav-header.
So I tried something else: Encoded a acelp.net-wma (8kHz; 8,5kbps), remuxed the wma to wav (wav play fine), and then encoded the the same file to "RealAudio 5.0 - 8,5kbps Voice". After that I wanted to replace the data in the wav with the one from the .ra, but when I looked at the files in the HexEditor I started laughing:
Data is bit-identical! The RealAudio Encoder 3.1 and Windows Media Encoder 7 produces identical output!
Real-Voice = WMA-Voice!!!!!
So it shouldn´t be a problem to decode sipr with acelp.net-decoder, also because both Encoders have the same bitrate modes (5kbps; 6,5kbps; 8,5kbps at 8kHz and 16kbps at 16kHz)

[haibane]

@gabest

The video freezing problem I am having only occurs with the mkv, however in rm, no matter how i drag the window or do other stuff, the video never freeze, it just start to drop frames. Besides, the video freeze occurs every time the cpu usage briefly reach 100%. It's not constanly at 100%, in fact, the cpu usage is at about average 30%. So, when exit from fullscreen, the video will freeze. This brings up another thing, when exit from fullscreen, the mkv version will eat up all the cpu compare to the rm version only use very little cpu.

About the sync problem, there are about 200-300ms difference uniformly through the whole clip between the audio and video, and the video did play vfr. I don't know if you really want the clips, because they are 100mb each........

[Stux]

Originally posted by karl_lillevold
This has been discussed a few times. All image processing in RV9 is inloop, thus non-separable and non-adjustable. The best known example of another codec having a similar feature, is H.264. The main advantage is that the codec gets to work with as clean frames as possible, increasing compression efficiency and video quality. The disadvantage is that it can not be adjusted without changing the bitstream, requiring the adjustment to be made by the encoder. This said, maybe one day for experimental purposes, I may make an adjustable encoder version, requiring a decoder update to go along with it.

3ivx D3.5 did this too ;)

[Sirber]

3ivx do inloop plus extreme post-processing? :confused:

[Dark-Cracker]

Hi,

i have made a little tool to directely use the dshow filter and connect the pins :) for the moment it only support .rmvb for input and create an .mkv for output. i will add support for .avi input and .mp3/.ogg/.aac/.srt/.ssa ... etc as soon i will find time to made some tests.

URL:
----

http://www.eclipsedvd.firstream.net/autodub/MatroskaMuxer.zip

how to use it :
---------------

1) be sure u have install the filters made by gabest.
2) launch the setup.bat (to registering the .dll)
3) start my tool select input .rmvb and output .mkv and wait.

WARNING :
---------
for the moment i don't have succeded to get the % of conversion u must wait your hard disk stop to work to know if the conversion is finish. Sorry i will try to find a better solution as soon as possible.

if u could post the result of your test, or bug report, it's the first time i play with dshow filter.

PS: very nice work gabest :)

Bye.

[Sirber]

Great!

Good job DC and Gabest! :D

[Animaniac]

Originally posted by Dark-Cracker
Hi,

i have made a little tool to directely use the dshow filter and connect the pins :) for the moment it only support .rmvb for input and create an .mkv for output. i will add support for .avi input and .mp3/.ogg/.aac/.srt/.ssa ... etc as soon i will find time to made some tests.

URL:
----

http://www.eclipsedvd.firstream.net/autodub/MatroskaMuxer.zip

how to use it :
---------------

1) be sure u have install the filters made by gabest.
2) launch the setup.bat (to registering the .dll)
3) start my tool select input .rmvb and output .mkv and wait.

WARNING :
---------
for the moment i don't have succeded to get the % of conversion u must wait your hard disk stop to work to know if the conversion is finish. Sorry i will try to find a better solution as soon as possible.

if u could post the result of your test, or bug report, it's the first time i play with dshow filter.

PS: very nice work gabest :)

Bye.

Here comes the Matroska complement to Koepi's OggMux!!

[Stux]

Originally posted by Sirber
3ivx do inloop plus extreme post-processing? :confused:

D1/2/3/3.5 did, D4 doesn't.

We found that doing in-loop filtering made it very hard to update the algorithms going forward

It was called "Artifact Reduced Encoding"

[faxmactor]

There was a request regarding Realmedia codecs in the beginning of this topic (RadicalEd, perhaps). After looooong experiments I managed to strip the necessary codecs from RealOne a few months ago. Now, combined with RealMediaSplitter (you are a awesome, Gabest!) it has more "style" (and smaller, too).

I know, that there may be more files than necessary. I dont have the necessary programming skills to determine the dependencies, so I was experiencing a lot.
It might be possible that not all the included DLLs are needed, but what I know is that lack of one single file can lead to rare, but ugly _really random_ crashes. :scared:

It WORKS perfectly! :D

If the moderators think so, I can post a direct or visit link here. Please notify me if I shouldn't do so.

[gabest]

S_O: I tried the sipro filter, but it wanted a nonstandard waveformatex structure and didn't play my files anyhow. But what is the problem with the real codec? To me it was working perfectly. Do you have sipr3260.dll at all? Also, can you upload a sample non-rm file with sipro audio? So that I can check the media type and compare to mine to correct it.

haibane: I'm sorry, I'm running out of ideas... Maybe with time this will be solved somehow.

faxmactor: Hehe, there aren't many dlls needed, and you can just dump them to system32 on a machine without realone, ... but I'm not encouraging anyone to do that of course.

[faxmactor]

Originally posted by gabest

faxmactor: Hehe, there aren't many dlls needed, and you can just dump them to system32 on a machine without realone, ... but I'm not encouraging anyone to do that of course.

(Under Win98SE, Hungarian) without the hole "Codecs", half the "Plugins" and one "Common" file, it was quite unstable. Perhaps it was MPC's fault :D Even a missing tfilesys.dll could have caused very random crashes (I had ripped all my hair one by one until I guessed ;)

Perhaps I am wrong (can you tell me what are really necessary from RealOne?). I have time, I'll try to reduce the surplus, but it is quite difficult, since the crashes were random. MPC also evolved since March, though, not to mention the standalone RealMediaSplitter now.

But anyway, the only ting I know that the present state is rock stable (as it can be under MS Windows :devil: )

[karl_lillevold]

RealMediaSplitter and its features are interesting topics, but i think it would perhaps be a good idea to not discuss what is related to illegal re-distribution of DLLs in this forum.

[kilg0r3]

1.)
why does is not work with ogm?

2.) Tried a stream consisting of one rmvb stream without audio plus one real audio stream muxed into mkv. Beside the already mentioned freezes after resizing the window. I encountered one instance where the video first started to stutter for one second and then stoped (video only). Furthermore the video also freezes when changing the focus on the windows. And, it never recovers on the next keyframe.
the matroska muxer and splitter used were Gabest's.

[midiguy]

Tested the latest filters. Real Media container works very well with fast seeking (although I haven't tested the audio), and MKV works, but every once in a while, the video freezes for 2 seconds or so, and then resumes playback.

[gabest]

Hm, I think I know why the mkv is like that... There is a larger buffer in the rm splitter, and because the audio frames are larger with realaudio more samples have to be buffered to allow an audio preload (the video renderer won't move until the audio renderer drives it with the clock, and this will block the splitter's video pin as well). Could you test the mkv again in graphedit, but so that you delete the audio filters after the splitter?

Btw, do you say the blade2 mkv sample plays badly for you? Because that should not suffer from this problem, it plays well at least on my machine.

[Assault]

@ Gabest

Hm...
The mkv sample doesn't freeze 1 or 2 seconds but when I play it with latest MPC the playback is a little bit jerky. It isn't very much but it isn't too hard to recognize.
But when I play the sample with Zoomplayer everything works fine. It doesn't play jerky at all. :confused:

My system:
Athlon 1,2 GHZ
256 mb RAM
Geforce4 TI 4200
WinXP, no SP

Assault