[FFmpeg-devel] Ffmpeg, UDP input source from satellite audio sync issues

Ibrahim Tachijian barhom at netsat.se
Thu Jun 21 10:34:20 EEST 2018

Hello Devin,

Thank you for all your input. I understand the A/V sync issues is a hard
task and we need to save the original video to catch where the issue

You may be well served to simultaneously write out a copy of the
> original TS to disk prior to transcoding.  This will allow you to
> debug cases where A/V sync was lost and reproduce the issue.  Also,
> it's much easier to be able to hand a sample to the developers and say
> "Try this 100MB sample and you can see for yourself" as opposed to not
> being able to provide any detail as to specifically what the source of
> the A/V sync issue was.

I will do my best to save the source and transcoded output to file to see
where the A/V sync started so that it can be easily reproduced.

> Also, bear in mind that the issue could be on the transcoding or the
> resulting decoder you are using to view the stream, and restarting the
> transcoder may cause the issue to "go away" in the decoder just
> because the stream is reset.  When dealing with A/V sync issues I
> commonly find two classes of issues:

Isn't the decoder considered "reset" everytime I re-launch VLC (if I see
sync issues in VLC) ? If yes then I can mostly rule out decoder issues.

I have another question for anyone who can answer it. I currently today do
a "ffprobe -i /tmp/hls/playlist.m3u8 -show_streams" and look at the audio
and videos "start_time" and I look at the difference.

I am currently considering a stream to be "out of sync" when the difference
is above 1.

What does difference of 1 in start_pts actually mean? Is it wrong or
correct to consider A/V out of sync when there is a difference in start_pts

*Ibrahim Tachijian*

CEO Net Sat AB
Mobile: +46723321377
barhom at netsat.se


*Net Sat ABC/O Jayway AB*Klara Ă–stra Kyrkogata 2B, 1TR, 111 52 Stockholm,
Office: +46 (0)8 408 394 53

More information about the ffmpeg-devel mailing list