In overlaying graphics/text onto a video stream, there is *no option* with
which to defer the start of the overlay clip.

There is also (AFAICT) *no video-file format* which includes transparency,
so the clip has to be provided as a humungous collection of PNG images
(short example here: https://www.youtube.com/watch?v=7aXnmuoiO4Q), it
cannot be turned into a video then the '-itsoffset' option used to bump it
along.

So why not use the 'fade' filter?

Every experiment thus far has had it ditching the alpha (transparency)
information, thus overlaying the main video stream with a black background,
then the actual overlay.

The movie filter has a 'seek_point' option which skips a certain portion of
the overlay when starting to write it atop the base video. One (I hope) of
two possible options would make this more useful. Each option would take a
single argument, being the start time within the base video at which the
overlaying begins.

This could be done either by simply not overlaying until that point, or by
repeatedly overlaying just the first frame of the overlay video (so in my
case, prepending a transparent PNG to the sequence would achieve that).

In the first case, an option entitled 'overlay_from' would sound about
right.

The second case is not so easy to name. Using 'repeat_until' describes
exactly what it does, however does not mention what it is repeating.
Morphing that into 'repeat_1st_until' overcomes that shortfall, yet sounds
less elegant.

Adding another option 'repeat_to_end' would simplify watermarking-type
operations, by watermarking up to the overlay start with the first frame,
then watermarking a section with the intervening frames of the overlay
movie, then watermarking from there to the end of the base video with the
final frame.

How say you...?

-- 
*Cheers — Leon*
_______________________________________________
libav-tools mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-tools

Reply via email to