On 5/7/17, Gyan <gyandoshi@gmail.com> wrote:

The video fade filter has a start_frame option which starts the fade effect when the ordinal count of the processed frame reaches the start_frame value e.g.

ffmpeg -f lavfi -i testsrc2=d=5 -vf trim=2,fade=s=25:d=1 vfade-works-by-frame-count.mp4

In the command above, the video output fades in starting at pts_time 3.00

whereas the command,

ffmpeg -f lavfi -i testsrc2=d=5 -vf trim=2,fade=st=1:d=1 vfade-fails-by-timestamp.mp4

fails because 'st' references timestamps and trim does not reset them.

With the afade filter, the command

ffmpeg -f lavfi -i sine=d=5 -af atrim=2,afade=ss=44100:d=1 afade-fails-by-sample-count.mp4

fails. Audio remains at native volume throughout.

But the command

ffmpeg -f lavfi -i sine=d=5 -af atrim=2,afade=ss=88200:d=1 afade-works-if-sample-count-is-pts.mp4

works. Audio fades in from pts_time 2.00 to 3.00. Which indicates that 'ss' refers to the original sample PTS rather than the ordinal count of processed audio samples.

Is this intended? If yes, can one emulate the behaviour of the video fade filter. If no, should I file a bug report?

Yes it is intended, If you want alternative fill feature request on out trac. Alternative you can reset pts with asetpts=.. filter.