I implemented something like this in a previous life. One thing I did was to insert JPEG comments into each frame containing an encoding marker (speex, etc) and

I implemented something like this in a previous life. One thing I did was to insert JPEG comments into each frame containing an encoding marker (speex, etc) and audio data. This allowed a standard consumer to see the MJPEG image as normal; but a modified client (including within a browser, using javascript) could decode the audio and video in sync and play back with sound.