Another awesome feature would be if the framebuffer was exposed. It is somewhat via canvas draw, but this isn't triggered by the video, so you will not get every single frame. If the video object had an "frameRendered" event we can use it to grab the pixels when available. One application would be seamless looping. If i could extract the first and last couple of frames, i could use those to hide the jank videos have when you seek to the start.