[ Thanks to An Anonymous
Reader for this link. ]
“The idea stems from the fact you don’t actually have to watch
every frame of a video playback to understand what is happening.
Instead, you can pull out frames of the video that are important to
the content’s progression. That includes the frames visible when
subtitles are displayed combined with “scene frames” which Mark
explains is the ‘midpoint of each scene.’ By combining the two you
get a collection of scene images that when played in sequence give
an accurate account of what is happening in any given video. You
view the key actions and you read the subtitles to form an almost
complete picture of what you’d get by watching the original video
instead.”