[FFmpeg-trac] #3678(avcodec:new): Regressions in the video duration/bitrate report

[FFmpeg]

#3678: Regressions in the video duration/bitrate report
------------------------------------+-----------------------------------
             Reporter:  Malizor     |                    Owner:
                 Type:  defect      |                   Status:  new
             Priority:  normal      |                Component:  avcodec
              Version:  git-master  |               Resolution:
             Keywords:              |               Blocked By:
             Blocking:              |  Reproduced by developer:  0
Analyzed by developer:  0           |
------------------------------------+-----------------------------------

Comment (by cehoyos):

 Replying to [comment:4 Malizor]:
 > Replying to [comment:3 cehoyos]:
 > > Replying to [comment:2 Malizor]:
 > >
 > > > The full video file is 40G and is 1h 26mn long.
 >
 > Sorry, in fact it is truly 1:48:31 long (as verified by playing the
 video in VLC 2.0.8)

 Could you add the exact file size?
 {{{ffmpeg -i input -f null -}}} tells you the exact length.

 > > > I understand what you say about the bitrate, but it's sure that it
 is truly higher than the detected 9118 kb/s.
 > >
 > > Of course.
 > > The bitrate of the file is unknown / wrong because the video stream
 doesn't store a bitrate.
 >
 > Apparently somthing changed in the way ffmpeg try to guess the bitrate,
 and it has regressed for some videos like this one.

 No. (Or at least I find this wording misleading.)
 The value written in mpegvideo streams is not the actual bitrate but the
 maximum bitrate.
 For your (mxf) file, the values are nearly identical, for (more typical)
 dvb streams, the value is completely wrong. It was therefore decided not
 to use the (possibly wrong) maximum bitrate value as actual video bitrate.

--
Ticket URL: <https://trac.ffmpeg.org/ticket/3678#comment:5>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker