From mike at haali.net Fri May 1 15:55:10 2009 From: mike at haali.net (=?UTF-8?B?0JzQuNGF0LDQuNC7INCc0LDRhtC90LXQsg==?=) Date: Fri, 01 May 2009 17:55:10 +0400 Subject: [Matroska-devel] Issue with seeking of .mkv files containing H.264 content In-Reply-To: <9D6A37B7E1AF664493D03F8D6109CC015B112C9895@NA-EXMSG-C102.redmond.corp.microsoft.com> References: <9D6A37B7E1AF664493D03F8D6109CC015B10FEE87B@NA-EXMSG-C102.redmond.corp.microsoft.com> <49F8145E.6000307@haali.net> <9D6A37B7E1AF664493D03F8D6109CC015B112C9895@NA-EXMSG-C102.redmond.corp.microsoft.com> Message-ID: <49FAFF3E.1000605@haali.net> Harsha Kikkeri (CODEC DSP) wrote: > We are seeing issues with seeking of .MKV files containing H.264 content. > After seeking, Haali source filter sends out a bunch of samples which have preroll=1(with negative timestamps, decoder ignores these timestamps) followed by samples with preroll=0. > Decoder decodes all these samples and only sends the samples with preroll=0 to downstream renderer. > When renderer receives these samples, timestamps on these samples are already behind the clock and this causes renderer to send late quality messages. > Haali source can check that the timestamp of the sample that it is sending out should be greater than current streamtime. > http://msdn2.microsoft.com/en-us/library/ms780225.aspx > > What we see, it that after a seek the timestamp of the first few samples (which have preroll=0) is behind current StreamTime. > This causes decoder to drop frames. This happens with ffdshow and Cyberlink decoders as well. What you describe is a situation when the seek is requested to a non-keyframe. My current code seeks to a previous keyframe and sends intermediate frames as preroll. Could you suggest a better algorithm to handle this case? I didn't find anything specific about it in DShow documentation. From shyams at microsoft.com Fri May 1 20:03:36 2009 From: shyams at microsoft.com (Shyam Sadhwani) Date: Fri, 1 May 2009 11:03:36 -0700 Subject: [Matroska-devel] Issue with seeking of .mkv files containing H.264 content In-Reply-To: <49FAFF3E.1000605@haali.net> References: <9D6A37B7E1AF664493D03F8D6109CC015B10FEE87B@NA-EXMSG-C102.redmond.corp.microsoft.com> <49F8145E.6000307@haali.net> <9D6A37B7E1AF664493D03F8D6109CC015B112C9895@NA-EXMSG-C102.redmond.corp.microsoft.com> <49FAFF3E.1000605@haali.net> Message-ID: <8CF87FB4831F784BA30F31982F8FC07B6224EAA534@NA-EXMSG-C116.redmond.corp.microsoft.com> Hi Mike, We do not want Haali source to change the seeking mechanism. Only issue is accuracy of timestamps on input samples after seek. We are fine with seeking to previous keyframe and send intermediate frames as preroll. The problem we are seeing is that after these intermediate preroll frames, the frames coming into decoder have timestamps (which are greater than 0) that are already behind current clock. I suspect Haali source is not setting correct timestamps after seek. Timestamps seem to become correct after next key frame most of the time. Thanks Shyam -----Original Message----- From: ?????? ?????? [mailto:mike at haali.net] Sent: Friday, May 01, 2009 6:55 AM To: Discussion about the current and future development of Matroska Cc: Yongjun Wu; Shyam Sadhwani Subject: Re: [Matroska-devel] Issue with seeking of .mkv files containing H.264 content Harsha Kikkeri (CODEC DSP) wrote: > We are seeing issues with seeking of .MKV files containing H.264 content. > After seeking, Haali source filter sends out a bunch of samples which have preroll=1(with negative timestamps, decoder ignores these timestamps) followed by samples with preroll=0. > Decoder decodes all these samples and only sends the samples with preroll=0 to downstream renderer. > When renderer receives these samples, timestamps on these samples are already behind the clock and this causes renderer to send late quality messages. > Haali source can check that the timestamp of the sample that it is sending out should be greater than current streamtime. > http://msdn2.microsoft.com/en-us/library/ms780225.aspx > > What we see, it that after a seek the timestamp of the first few samples (which have preroll=0) is behind current StreamTime. > This causes decoder to drop frames. This happens with ffdshow and Cyberlink decoders as well. What you describe is a situation when the seek is requested to a non-keyframe. My current code seeks to a previous keyframe and sends intermediate frames as preroll. Could you suggest a better algorithm to handle this case? I didn't find anything specific about it in DShow documentation. From chris at wiesneronline.net Wed May 6 20:20:48 2009 From: chris at wiesneronline.net (Christian Wiesner) Date: Wed, 6 May 2009 20:20:48 +0200 Subject: [Matroska-devel] Parsing MKV headers in Java In-Reply-To: <4891aa620905060552g4375171et49c2afc8d3ab10cd@mail.gmail.com> References: <4891aa620905060552g4375171et49c2afc8d3ab10cd@mail.gmail.com> Message-ID: <264202f40905061120o40439b8eoa65066fcafde65c1@mail.gmail.com> Of course ! Which language would you prefer ? Christian 2009/5/6 Marco > Hi, I read your message on the abcAVI mailing list > > > Hi, > > > > are there anyplans to extend support for this plugin also to MKV > > (matroska) files ? We can give you free, complete libraries in either > > C, C++, Java etc. to allow to parse the file headers of MKV and MKA > > (audio) files easily. > > > > If you are interested, contact us at matroska-de... at lists.matroska.org > > > > Christian > > matroska project admin > > http://www.matroska.org > > > Is there a free library to parse headers of mkv files (to gather the > information > that mkvinfo displays)? > > many thanks, > marco paolieri > (Italy) > -------------- next part -------------- An HTML attachment was scrubbed... URL: From michele.pradella at selea.com Thu May 7 10:44:49 2009 From: michele.pradella at selea.com (Michele Pradella) Date: Thu, 07 May 2009 10:44:49 +0200 Subject: [Matroska-devel] Muxer Demuxer Message-ID: <4A029F81.8050407@selea.com> Hi, I would like to implement in my application the muxer/demuxer of MKV format, to save and read that kind of container. I have downloaded *libmatroska* and *libebml *but I can't find how to use this library. Is there a wiki or a documented api, or an example on how to use it? best regards, Michele Pradella From remco at qubis.nl Tue May 12 00:03:24 2009 From: remco at qubis.nl (Remco Bloemen) Date: Tue, 12 May 2009 00:03:24 +0200 Subject: [Matroska-devel] Proposal for future EBML Message-ID: <200905120003.24686.Remco@qubis.nl> Dear Matroska and ebml developers, First of all, thank you for this sensible and efficient standard! It is just wonderful how the ebml element overhead can be as low as the size of an '<' character in an utf16 encoded xml document :) About three years ago there where some messages in the mailing list about a future ebml specification which would include things like namespaces and a schema. I am working on a concrete proposal for such an extension. The draft proposal can be found at: http://www.qubis.nl/ebml/Ebmlv2.html I am also working on a lean C# implementation of this proposal which I will release when it is ready. I might port it to C++ should this be useful. Main Question: Is there still interest in a future version of ebml? If so, I would appreciate it if you review my proposal. Should a rationale be missing or unclear, please ask me. Should it be of interest, I intend to use this slightly improved ebml as a means of data storage and network communication in some future projects of Qubis. Kind regards, Remco Bloemen From rodrigo.madera at gmail.com Tue May 12 00:34:19 2009 From: rodrigo.madera at gmail.com (Rodrigo Madera) Date: Mon, 11 May 2009 19:34:19 -0300 Subject: [Matroska-devel] Proposal for future EBML In-Reply-To: <200905120003.24686.Remco@qubis.nl> References: <200905120003.24686.Remco@qubis.nl> Message-ID: <3cf983d0905111534w44366d4cw25044814c8188b2a@mail.gmail.com> I started reading your document. I read: Note on compatibility: This specification is breaks binary compatibility with the existing ebml specification... I immediately closed your website. Regards, Rodrigo On Mon, May 11, 2009 at 7:03 PM, Remco Bloemen wrote: > Dear Matroska and ebml developers, > > First of all, thank you for this sensible and efficient standard! It is > just > wonderful how the ebml element overhead can be as low as the size of an '<' > character in an utf16 encoded xml document :) > > About three years ago there where some messages in the mailing list about a > future ebml specification which would include things like namespaces and a > schema. I am working on a concrete proposal for such an extension. The > draft > proposal can be found at: > > http://www.qubis.nl/ebml/Ebmlv2.html > > I am also working on a lean C# implementation of this proposal which I will > release when it is ready. I might port it to C++ should this be useful. > > Main Question: Is there still interest in a future version of ebml? If so, > I > would appreciate it if you review my proposal. Should a rationale be > missing > or unclear, please ask me. > > Should it be of interest, I intend to use this slightly improved ebml as a > means of data storage and network communication in some future projects of > Qubis. > > > Kind regards, > > Remco Bloemen > > _______________________________________________ > Matroska-devel mailing list > Matroska-devel at lists.matroska.org > http://lists.matroska.org/cgi-bin/mailman/listinfo/matroska-devel > Read Matroska-Devel on GMane: > http://dir.gmane.org/gmane.comp.multimedia.matroska.devel > -------------- next part -------------- An HTML attachment was scrubbed... URL: From remco at qubis.nl Tue May 12 07:57:13 2009 From: remco at qubis.nl (Remco Bloemen) Date: Tue, 12 May 2009 07:57:13 +0200 Subject: [Matroska-devel] Proposal for future EBML In-Reply-To: <3cf983d0905111534w44366d4cw25044814c8188b2a@mail.gmail.com> References: <200905120003.24686.Remco@qubis.nl> <3cf983d0905111534w44366d4cw25044814c8188b2a@mail.gmail.com> Message-ID: <200905120757.13312.Remco@qubis.nl> On Tuesday 12 May 2009 00:34:19 Rodrigo Madera wrote: > I started reading your document. I read: > Note on compatibility: This specification is breaks binary compatibility > with the existing ebml specification... ..., this is a necesary consequence of introducing namespaces. However ebml v1 is syntactically valid ebml v2. So all current ebml is forward compatible with this proposal and can be read by syntax-level parsers implementing this proposal. To allow a semantic-level parser to be backwards compatible it is only necessary to mix the DocType namespace in the ebml namespace. This was discussed in (among others): http://lists.matroska.org/pipermail/matroska-devel/2006-April/002906.html > I immediately closed your website. I am very sorry to hear that. Good thing the notice is right at the top then :) Kind Regards, Remco Bloemen > Regards, > Rodrigo > > On Mon, May 11, 2009 at 7:03 PM, Remco Bloemen wrote: > > Dear Matroska and ebml developers, > > > > First of all, thank you for this sensible and efficient standard! It is > > just > > wonderful how the ebml element overhead can be as low as the size of an > > '<' character in an utf16 encoded xml document :) > > > > About three years ago there where some messages in the mailing list about > > a future ebml specification which would include things like namespaces > > and a schema. I am working on a concrete proposal for such an extension. > > The draft > > proposal can be found at: > > > > http://www.qubis.nl/ebml/Ebmlv2.html > > > > I am also working on a lean C# implementation of this proposal which I > > will release when it is ready. I might port it to C++ should this be > > useful. > > > > Main Question: Is there still interest in a future version of ebml? If > > so, I > > would appreciate it if you review my proposal. Should a rationale be > > missing > > or unclear, please ask me. > > > > Should it be of interest, I intend to use this slightly improved ebml as > > a means of data storage and network communication in some future projects > > of Qubis. > > > > > > Kind regards, > > > > Remco Bloemen > > > > _______________________________________________ > > Matroska-devel mailing list > > Matroska-devel at lists.matroska.org > > http://lists.matroska.org/cgi-bin/mailman/listinfo/matroska-devel > > Read Matroska-Devel on GMane: > > http://dir.gmane.org/gmane.comp.multimedia.matroska.devel From brbromo at gmail.com Thu May 14 22:26:45 2009 From: brbromo at gmail.com (Marco) Date: Thu, 14 May 2009 22:26:45 +0200 Subject: [Matroska-devel] Parsing MKV headers in Java In-Reply-To: <4891aa620905140708t2fe53c9ag39d168c4416f53c9@mail.gmail.com> References: <4891aa620905060552g4375171et49c2afc8d3ab10cd@mail.gmail.com> <264202f40905061120o40439b8eoa65066fcafde65c1@mail.gmail.com> <4891aa620905140708t2fe53c9ag39d168c4416f53c9@mail.gmail.com> Message-ID: <4891aa620905141326u7e44308csa6dc5acc202163bb@mail.gmail.com> 2009/5/6 Christian Wiesner > > > 2009/5/6 Marco >> >> Is there a free library to parse headers of mkv files (to gather the >> information >> that mkvinfo displays)? >> >> > Of course ! > > Which language would you prefer ? > > Christian > > Java!! many thanks! marco -------------- next part -------------- An HTML attachment was scrubbed... URL: From nunnaramakrishna at gmail.com Tue May 19 14:57:06 2009 From: nunnaramakrishna at gmail.com (ramakrishna nunna) Date: Tue, 19 May 2009 18:27:06 +0530 Subject: [Matroska-devel] Streaming in matroska Message-ID: <9860923e0905190557h12f0764dy9051885813d25963@mail.gmail.com> Hi, I, Ramakrishna started implementation of Matroska Parser. I have some doubts regarding streaming functionality in matroska. Unlike OGG, in matroska developers didnot put any limits on the size of the cluster. So, for implementation of streaming functionality how we have to take inputs to the Parser? Was it cluster by cluster or blockgroup by blockgroup? If it is cluster by cluster, then in real time scenario is it possible to have that much bigger buffers to hold cluster data? Why the developers didnot kept cluster size limitation such page structure in OGG, else it would have been much easier for streaming functionality? Here I don't want to comment about the design. But, I want to get some inputs on my query. Thanks in advance. -- Regards, Ramakrishna N ------------------------------------------------------------------------------ Engineer, Media Processing Ittiam Systems | www.ittiam.com A Red Herring Asia Top 100 Company, 2005 Forward Concepts' Worlds Most Preferred DSP IP Supplier, 2004, 2005, 2006,2007 -------------- next part -------------- An HTML attachment was scrubbed... URL: From dlmcpaul at gmail.com Wed May 20 07:43:08 2009 From: dlmcpaul at gmail.com (David McPaul) Date: Wed, 20 May 2009 15:43:08 +1000 Subject: [Matroska-devel] libebml and libmatroska Message-ID: Hello, I have used these 2 libraries to add matroska support to the Haiku OS project (http://www.haiku-os.org/) I have recieved some bug reports about matroska v2 files not working. Are these libraries still under development? Are there plans to update these libraries to handle V2 matroska? -- Cheers David From dellagustin at gmail.com Wed May 20 19:21:21 2009 From: dellagustin at gmail.com (Guilherme Dellagustin) Date: Wed, 20 May 2009 14:21:21 -0300 Subject: [Matroska-devel] TRP Splitting and Problem with TS with AAC audio Message-ID: Hi there! thanks for this useful tool. I'm from a group that is trying to play files recorded from a HDTV (ISDB-Tb, used in Brasil and probably ind the rest of Latin America in the future) receiver and PVR, the files are stored in TRP extension, wich we can transform into TS using H264TS Cutter. The problem here is that the audio is encoded using AAC LC, but the splitter sends MPEG1AudioPayload to the audio decoder, and it obviously do not work. So, it would be a good thing if in the future Haali Media Splitter could work with trp and ts using AAC audio. If you have any sugestion to solve this by now, it would be very welcome! Thanks Again. -------------- next part -------------- An HTML attachment was scrubbed... URL: From chris at matroska.org Sat May 30 12:33:41 2009 From: chris at matroska.org (Christian Wiesner) Date: Sat, 30 May 2009 12:33:41 +0200 Subject: [Matroska-devel] Streaming in matroska In-Reply-To: <9860923e0905190557h12f0764dy9051885813d25963@mail.gmail.com> References: <9860923e0905190557h12f0764dy9051885813d25963@mail.gmail.com> Message-ID: <21e2294b0905300333sdddead9uf5bfb29fcac51de9@mail.gmail.com> The focus for the development of matroska, originally, was not put on streaming. OGG was available for that already, and there was enough bad feelings between the two teams that made it necessary to clearly differentiate the purposes of both containers. However, there has been a lot of discussion lateron about a specific 'streaming mode' for matroska, you can read all this in the mailing list archives on gmane.org , mainly in matroska-devel . Needless to say, nobody ever had the time to really look in this new extension of the existing specs, so it never happened. One reason is certainly that there are enough working streaming formats existing, and the big Forte of matroska, its ability to use many different audio and video codecs in a MKV file, for streaming purposes is rather a disadvantage than really a benefit .... Regards Christian matroska project admin 2009/5/19 ramakrishna nunna > Hi, > I, Ramakrishna started implementation of Matroska Parser. I have some > doubts regarding streaming functionality in matroska. > Unlike OGG, in matroska developers didnot put any limits on the size of the > cluster. So, for implementation of streaming functionality > how we have to take inputs to the Parser? Was it cluster by cluster or > blockgroup by blockgroup? > If it is cluster by cluster, then in real time scenario is it possible to > have that much bigger buffers to hold cluster data? > Why the developers didnot kept cluster size limitation such page structure > in OGG, else it would have been much easier for > streaming functionality? > Here I don't want to comment about the design. But, I want to get some > inputs on my query. > > Thanks in advance. > > -- > Regards, > Ramakrishna N > > ------------------------------------------------------------------------------ > Engineer, Media Processing > Ittiam Systems | www.ittiam.com > A Red Herring Asia Top 100 Company, 2005 > Forward Concepts' Worlds Most Preferred DSP IP Supplier, 2004, 2005, > 2006,2007 > > _______________________________________________ > Matroska-devel mailing list > Matroska-devel at lists.matroska.org > http://lists.matroska.org/cgi-bin/mailman/listinfo/matroska-devel > Read Matroska-Devel on GMane: > http://dir.gmane.org/gmane.comp.multimedia.matroska.devel > -------------- next part -------------- An HTML attachment was scrubbed... URL: