I've ported two operations from gggl to GEGL now. Input and output operations using ffmpeg's libavcodec and libavformat. The code basically works and provides the basic infrastructure to make a video compositing/editing app. Some remarks about the state of this code:

ff-load: just like an image input loader except it also takes the frame number desired. It currently uses a very old-style ffmpeg API for decoding, it should probably be updated/simplified according to the following document at some point:

http://www.inb.uni-luebeck.de/~boehme/libavcodec_update.html

ff-load currently does a very inefficient form of seeking, it always decodes from the previously decoded frame number to the desired frame number, and if the frame number is smaller than the current decode position in the video file it will start decoding from scratch. It seems like ffmpeg provides the needed infrastructure to aid a frame accurate random access that should be quite a bit faster but at least we've got a proof of concept implementation that allows doing some fun things with video.

ff-save includes the basic infrastructure to mix audio into the output stream as well but this code is currently completely unused.

There is need for some code cleanup as well as checking of the memory handling in both these plug-ins but I've committed them to the workshop (the workshop operations are operations that are not built by default), and successfully applied effect filters and applied overlays to a video with a small script from rgegl.

/Øyvind K.