I dont know about ffmpeg, but libtheora allows you to specify the number of threads to do video frame decoding - essentially, there is a pool of decoded read-ahead frames, and a specific number of threads fill a pool buffer, which are consumed by the player. Libtheora has found a sweet spot of 15 read-ahead frames. For a HD project I worked on (libtheora fills OpenGL texture buffer), 2 threads were more than adequate at decoding 1920x1080 video without dropped frames. As a mater of fact, the CPU time spent on decoding the Theora encoded video (30 frames per second) was so low, that I was still able to get >3000 fps consistantly with VSYNC off (obviously the commercial product has VSync On). Decoding is not a problem (for Theora) - but with newer codecs, things may change. Zenja On Tue, Mar 9, 2010 at 2:36 PM, François Revol <revol@xxxxxxx> wrote: > Now about what to show, dunno, you might have a look at ffmpeg, they > have some support for threaded video decoding, though I'm not sure our > Media Kit wrapper supports it yet. > > François. > >