[theora] Encoding vorbis with theora

[Gregory Maxwell]

On Sun, Jan 16, 2011 at 10:35 AM, Stuart Fisher <stuart.fisher at fig7.com> wrote:
> Hi Ernie,
>
> I was thinking about similar issues some time ago. I'd been writing some
> code to playback theora clips.
>
> The way I see it is that a player has to assume the audio track is always
> a continuous audio stream. The video track, on the other hand, doesn't have
> to be. The reason is that in order to playback the clip you need to have
> synchronisation between the audio and video. The audio stream provides that
> synchronisation (the player uses the audio clock to determine when video
> frames should be displayed). So in other words I think you need to encode
> the audio as a continuous stream (i.e. just encode silence between
> the sections of audio).
>
> That's the way I figured it out in my mind, but I wonder if there is a
> defnitive explanation somewhere of how theora encoders and decoders should
> handle discontinuity in the timeline.


Both Theora and Vorbis are continuous stream types. Kate is an example
of a discontinuous stream.
(http://xiph.org/ogg/doc/ogg-multiplex.html)

Both formats have efficient ways of coding inactivity (digital silence
in the case of vorbis, no motion in the case of theora), so you should
just feed them inactivity when you want them to be inactive.