Instead of rendering the complete video, this would store the information that created the video in various compressed flows, that are opened to editing, and quick rendering on the fly.

Users can change the texts, images and effects that build the video, and then watch their video in the same way as "preview" mode for video editing software.

The 'Publish' action, only "compiles" and compresses the information, which is still mostly separated per frame and media, or at least can be retrieved as such, and is rendered then rendered in this (open source? HTML6?) video player, when 'played'.

It will include vector animation, video effects, image layers and layer interactions, frame transitions and video tracks. Optimization could pull only the needed information during a reconstruction and playback, so muted tracks for instance would not need to be downloaded, as well as faded off parts.
-- pashute, Jul 27 2014

random, halfbakery