Re: [FFmpeg-devel] [PATCH 4/6] doc/muxers: add mkvtimestamp_v2

[Stefano Sabatini <stefasab@gmail.com>]

On date Tuesday 2024-04-16 20:09:19 +0200, Andreas Rheinhardt wrote:
> Stefano Sabatini:
> > On date Tuesday 2024-04-16 12:50:19 +0200, Andreas Rheinhardt wrote:
> >> Stefano Sabatini:
> >>> ---
> >>>  doc/muxers.texi | 8 ++++++++
> >>>  1 file changed, 8 insertions(+)
> >>>
> >>> diff --git a/doc/muxers.texi b/doc/muxers.texi
> >>> index f94513527d..490d5557bf 100644
> >>> --- a/doc/muxers.texi
> >>> +++ b/doc/muxers.texi
> >>> @@ -2933,6 +2933,14 @@ MicroDVD subtitle format muxer.
> >>>  
> >>>  This muxer accepts a single @samp{microdvd} subtitles stream.
> >>>  
> >>> +@section mkvtimestamp_v2
> >>> +mkvtoolnix v2 timecode format muxer.
> >>> +
> >>> +Write the PTS rawvideo frame to the output, as supported by the
> >>> +@command{mkvextact} tool from the @command{mkvtoolnix} suite.
> >>> +
> >>> +This muxer accepts a single @samp{rawvideo} stream.
> >>> +
> >>>  @section mp3
> >>>  
> >>>  The MP3 muxer writes a raw MP3 stream with the following optional features:
> >>
> > 
> >> This is wrong: MKVToolNix switched to "# timestamp format v2" a long
> >> time ago (we still write the old "# timecode format v2" header);
> >> furthermore, MKVToolNix actually uses pts (which it reorders to be
> >> ascending), not dts like our muxer. Furthermore MKVToolNix does not
> >> force a 1ms precision on timestamps.
> > 
> > Correct.
> > 
> > I compared the output of the muxer and of mkvtoolnix extract
> > timestamp_v2 and I'm not yet clear about the timestamp differences I'm
> > observing (the muxer output maps with the timestamps, the mkvtoolnix
> > timestamps differ by a few ms). But I think also mkvtoolnix use a 1ms
> > timebase.
> 
> The accuracy of the timestamps output by mkvextract is determined by the
> TimestampScale of the file in question; it is most often 1ms when the
> file has video.

> You need to provide more details if you want these discrepancies to be
> analyzed.

Probably not worth the effort.

> > Also, IIRC there is no generic way to reorder PTSs, so this might
> > account for another difference which might be difficult to implement
> > generically.
> 

> Write them into a buffer and reorder them at the end?
> (No, I have no intention to actually implement this. I am rather leaning
> to "this muxer should not exist".)

I also think we have better tools at this point (one being ffprobe
-show_packets) but we should not drop it before deprecating it.

Plan: av_tree to insert elements in a constant-size buffer or store in
a buffer sorted once at the end. We probably should skip PTS=NA
elements.

Dropping the doc patch as the implementation is broken. 

Will apply the rest of the patchset soon.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".