[FFmpeg-user] ffplay/mpv/mplayer large aac/adts seek regression.

Andy Furniss adf.lists at gmail.com
Fri Jan 1 15:15:54 CET 2016

Andy Furniss wrote:
> Doesn't affect ffmpeg -ss but interactive seeking on very large aac
> (adts) files is way out with current ffplay, mpv and mplayer.
> Pressing "up" to skip forward a minute ends up at 1:37:xx on one
> sample.
> Making the samples smaller reduces the effect so it wouldn't be seen
> on a "normal" length sample. The samples I am testing are half gig
> 3-4 hour radio recordings.

More playing - it's not just aac. I tried with some generated files and
it's reproduceable with those if long enough.

Noticed some wierds with aac encoder on the way.

ffm -f lavfi -i "sine=frequency=800:duration=7200" -c:a aac -b:a 320k 

works and produces seek issue but is very slow - speed = 4x and not 320k.

Omitting the -b:a 320k speed is 24x, still slowish.

Try above with both aac replaced by ac3 speed= 456x bitrate is 320k but 
no seek issue.

Repeat ac3 with duration=14400 and the seek issue is there.

More information about the ffmpeg-user mailing list