[Ffmpeg-devel] ScreenKast exporting

[Karl H. Beckers]

Hi all,
isn't the summary of this kinda:

Bram wants to use ffmpeg to encode but then only hand the bits that
change to the encoder. If I'm not completely mistaken, that's the major
task of most encoders, finding out what it is that has changed and then
only storing the deltas. Hence it sounds like what Bram's asking for is
a way to tell the encoder "only look at this or that part of the frame"
which is like encoding the input to the encoder, which again sounds like
it can only make a lot of sense if you can find out what changed more
quickly than the encoder could, which is prolly only the case if you
KNOW what's changed without having to determine it or if you have found
a clever motion detection algorithm. And then you could just add that to
a codec, no?

I'm in a similar situation, and if I could know what changed I wouldn't
need to get a complete new frame, but capture a potentially much smaller
part and insert it into the old frame, but I can't see how you would
determine what's changed in the new frame without looking at it (and
therefore getting it first).

Karl.