The original library I was experimenting with was from Main Concept. Basically we were just decoding the frame to it's raw format, reading out the bytes/pixels of two frames side by side with some black pixels in between, and then trying to encode this new frame. I hope there's an easier, more widely applicable way.