i still want to find a way to do this with ffmpeg (for the video) and sox (for the audio) plus maybe slide changeover detection, slide text recognition, face detection (to remove motion from speaker) etc etc etc. Feels like fixed camera video is not too hard to process...