[FFmpeg-devel] About guess_correct_pts / AVFrame.best_effort_timestamp

Baptiste Coudurier baptiste.coudurier
Wed Feb 16 21:48:11 CET 2011

On 2/16/11 5:45 AM, M?ns Rullg?rd wrote:
> Luca Barbato <lu_zero at gentoo.org> writes:
>> On 02/16/2011 01:26 PM, Nicolas George wrote:
>>> 2. A field in AVFrame, set at the same time as the rest of the structure.
>>>    Exporting the underlying logic is ok, but only if there is a need for it.
>>> 3. The current algorithm seems fine to me for a start,
>>>    but I do not insist on it if someone proposes something better.
>>>    There have been hints of problems in the timestamp logic, but nothing
>>>    precise that could be actually pinned on this algorithm.
>> The current algorithm and all the heuristics present in libavfilter
>> should be consolidated and made optional. We have too many workarounds
>> that might solve something breaking something else (e.g. the mpegts
>> produced by some old vlc fed by an axis camera are wrongly workarounded
>> as I recently discovered).
> More importantly, it breaks h264 in mkv, despite the container being
> complete and correct.

Complete as in no dts ?
and correct as in most muxers uses 1/1000 as a timebase ?

Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
FFmpeg maintainer                                  http://www.ffmpeg.org

More information about the ffmpeg-devel mailing list