[mythtv-commits] Ticket #1853: svn-0.19-fixes MPEG2->MPEG4 transcoding corrupts some recordings ("green squiggly mess").
MythTV
mythtv at cvs.mythtv.org
Wed May 24 09:46:52 EDT 2006
#1853: svn-0.19-fixes MPEG2->MPEG4 transcoding corrupts some recordings ("green
squiggly mess").
-----------------------+----------------------------------------------------
Reporter: oa at iki.fi | Owner: ijr
Type: defect | Status: new
Priority: minor | Milestone:
Component: mythtv | Version:
Severity: medium |
-----------------------+----------------------------------------------------
DVB-T recordings get frequently corrupted if transcoded to MPEG4,
apparently without regard to exact transcoding profile settings. This
seems to be caused by the MPEG2 decoder rather than the MPEG4 encoder, and
may to be related to aspect ratio changes or other stream disruptions as
it usually is triggered by a change into/out of/between commercials.
A 40MB example of a beginning of a recording which demonstrates this is at
http://www.fishpool.org/misc/mythtv-dvb-capture-fails-libmpeg2.mpg.
Additionally, I have found that this file plays correctly (prior to
transcoding) in MythTV when using the "Standard" (ffmpeg) decoder, fails
with the same symptoms with the "Libmpeg2" decoder, and exhibits spurious,
minor blocking with the "Standard XvMC" decoder. The same behavior can
also be demonstrated with "mplayer -vfm ffmpeg" vs "mplayer -vfm
libmpeg2".
As further evidence that the error probably is in the decoder, I've seen
the problem with various transcoder bitrates, HQ, 4MV and deinterlacing
options on and off, and with or without rescaling of the frame size. What
I can't understand, however, is that to the best of my understanding,
mythtranscode should be using ffmpeg's decoder, and I've NEVER encountered
the problem with untranscoded recordings when using the Standard playback
decoder.
As far as the test material, you won't see the error at the immediate
beginning of the recording (though on some recordings it _does_ happen
immediately), but watch through the first minutes, and it should happen
reliably. The problem has been in 0.19-fixes for a long time, and is
demonstrable with at least up to 9850.
Discussion on the -users list: http://mythtv.org/pipermail/mythtv-users
/2006-May/136019.html
--
Ticket URL: <http://cvs.mythtv.org/trac/ticket/1853>
MythTV <http://www.mythtv.org/>
MythTV
More information about the mythtv-commits
mailing list