Multiple input files "wildcard" feature request
(too old to reply)
Leon Brooks
2011-08-25 06:16:03 UTC
Raw Message
Using "thingy%03d.jpg" to read a whole pile of images works well, now I'd like
to get all ambitious by specifying something along these lines:


"Regexp" would be any standard regular expression, immediately carrying three
benefits: [1] specifying a sequence of images with (an) interruption(s) would
produce the generally-expected result of including all of the matched images,
regardless of numerical gaps; & [2] the images would not need to be numerically
named at all; & [3] specifying a subset of the available images would be much

The trailing 'r' is conceptually optional, however advantages in requiring it
include the ability to do things like prefix it with 'i' to make the matching
case-independent, make it 'R' to reverse the order in which matches are
processed, prefix it with 'd' or '/' to prompt recursive file-searching, etc.

It would be convenient (not necessary) to be able surround the regexp with
{braces} or (parentheses) rather than [brackets] as this would permit more
flexibility in specifying a regexp from whatever application (or shell) spawned
the ffmpeg instance.

It would be Unixly programmatic to be able to specify (subexpressions) within
the regexp & deploy those in settings (so 'frog%[.*\(\-[0-9]*\)\.jpg' would
match frogABC-12.jpg then frogPQR-8.jpg then frogXYZZY-27.jpg, making the
parameter -12, -8 or -27 available as a frame-time in seconds... wrap your head
around this idea:

ffmpeg '-r%1/(0-\1)' 'frog%[.*\(\-[0-9]*\)\.jpg' -target ntsc-dvd -y mydvd.mpg

Maybe optionally embed a Ruby interpreter into ffmpeg? (-:

Sneaking a second feature request in: for sequences of images, would it be
possible to specify a fade-in-time (during which this image is matted in front
of the previous image, with opacity linearly ascending from 0% to 100%),
show-time, then fade-out-time (during which the opacity is reduced linearly from
100% to 0% with the next image matted behind)? In conflict, either fade-out time
rules or use the mean of the two times.

Revision-point on that: how much harder would it be to specify a 'tween image,
which is faded in after & faded out before each image in the normally-specified

Shutting up now on the topic of specifying an opacity template image... (-: