What should a three-minute automatic video recap preserve from an hour of footage?

An automatic video recap can start with a game, interview, documentary, vlog, or film. The detail that carries the story changes with the source. In a sports recap it might be a defensive move that opens a pass; in an interview it might be the qualification that makes a quote accurate.

Take an hour of footage and a three-minute target for a first-time viewer. My first requirement would be to list the moments that change what happens next, then point the editor to the visual or spoken evidence for each. In the fictional film example in the diagram, a shot of the mark on a key explains why the detective chooses that door.

I would ask the automatic editor for a first version with editable shots, narration and captions. Then I’d ask a reviewer why the next action happened. If their answer depends on seeing the original, restore the missing evidence or adjust its timing.

For makers who ship video tools, which input would you ask editors to supply first: must-keep story beats, source timestamps, or a short note explaining the audience? What would you leave for review after the first edit?