[Theora-dev] Theora and Vorbis problems
illiminable
ogg at illiminable.com
Wed May 24 12:12:13 PDT 2006
Hi,
While it's true your file seems valid spec-wise, it suffers from being
poorly multiplexed... ie coarse muxing.
This is the first video page
--------------------------------------
Page Location : 73374 to 77589
Ver No : 0
Head Flags :
Granule Pos : 254
Serial No : 1792863441
Seq No : 2
Checksum : 279992667
Num Segs : 255
------------------------
Head Size : 282
Data Size : 3933
Page Size : 4215
Num Packets : 255
--------------------------------------
There's about 20 pages of audio before the first video page, and the first
video page contains 255 frames. This is due to the way ogg sorts pages by
end-time, it causes pathological files to be created with the default paging
algorithm particularly when the size of the video packets is small. VLC will
probably play, but most others things probably won't. My directshow code has
will only let there be about a 75 packet delta between current play time
between streams. That's typically about 3 seconds, which is already pretty
excessive. Everything could just make their buffers even bigger... but my
position is that these files are broken and it's better to discourage people
from making them, rather than making stupidly large buffers. Also files like
this will cause problems as more embedded devices start to use theora.
Basically the file is "ten seconds of audio data" then "ten seconds of
video data" rather than being nicely interleaved. When you mux, particularly
when you mux "easy" video such as this one, you have to make sure you flush
the stream after some number of packets ~50 say, or some amount of time say
1-2 seconds if more than that would normally continue to fit in one page. I
think ffmpeg2theora is the only tool currently that properly does this.
Cheers,
Zen.
----- Original Message -----
From: "Philip Heron" <phil at firestorm.cx>
To: "Theora Development" <theora-dev at xiph.org>
Sent: Thursday, May 25, 2006 1:29 AM
Subject: [Theora-dev] Theora and Vorbis problems
> _______________________________________________
> Theora-dev mailing list
> Theora-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/theora-dev
>
More information about the Theora-dev
mailing list