[Matroska-devel] Haali parser: Suggestions for Audio Stream Switcher

Steve Lhomme steve.lhomme at free.fr
Mon Jan 3 13:45:32 CET 2005

Алексей Сурженко a écrit :
> Dear Matroska-developers, dear Mikhail Matsnev!

Hi Алексей,

> Although some people (including me) long awaited this feature, i was a bit disappointed by its realization. What do i mean? actually, i often mixed mkv-movies with several audio streams, for example,
> Stream 1 - original audiotrack (ogg, 48KHz, stereo, avg. bitrate of appr. 120-130kbps)
> Stream 2 - the dialog translation by a lector (voice only - ogg, 22KHz, mono, avg. bitrate of 10-20kbps)
> Stream 3 - my own comments about the movie mistakes (ogg, 22KHz, mono, avg. bitrate of 10-20kbps)

Great job. I'm glad advanced users like you find what they need in 
Matroska :)

> Certainly, Speex would be more perfect choice for Streams 2 & 3, but mkvtoolnix still does not allow me to mux Speex-streams... Anyway, ogg is also quite good in the speech encoding.

Hopefully that shouldn't be too hard to add.

> So, my proprosition is either
> 1) to add an opportunity to disable the audioswitcher 

Probably quite complicated given the current design (which I haven't 
looked at closely). But allowing 0 to infinite audio streams should be 
done one day. I'm sure some people would have use for not decoding any 
audio at all in some movies.

> 2) or, what is even better, to make the audio selection following the principle "any combination of the embedded streams", but not "only one of them".

This is almost the same problem as above. Because AFAIK this splitter 
reconfigures the audio pins whenever another stream is selected. That 
means it's either 0 if the segment has no audio or 1 if it has plenty. 
But no new pin is created on the fly. It may lead to some problems with 
some players but I'm not sure.

For sure that would be the nicest way of doing things. Selecting the 
stream (on/off) from the stream list...

I also noticed this problem when playing a file in TCMP. Neither the 
audio or subtitle switching were available to TCMP, only from the filter 
(and sometimes it would freeze after the switch).

> ii) when the tracks are muxed in the manner i wrote, the original soundtrack remains untouched - it might be useful both for playing movie back without the lectors comments (for those who know English well or study this language) or for editing and/or replacing the lector voice by any other local language (Russian, German, French)...

It could also be used for multi-track audio where you could 
enable/disable some instruments.

> Although the same feature could be achieved by muxing the lector voice as a separate track in DolbySurround-tracks, i do believe that changing a bit alternative MKV-parser could solve this problem in the most reasonable manner...


> Faithfully, 
> Aleksei Surzhenko 
> (aka NuPogodi)

Thanks for your positive contribution !

More information about the Matroska-devel mailing list