Skip to main content

Why the editor watches the footage before it cuts

2 min read

Archive footage is messy. A four-minute newsreel about a launch might have twenty seconds of the launch. The rest is a title card, a slate, a man at a desk, a crowd, and a few frames of black where the film was spliced. An editor who drops the whole reel on the timeline gets all of that on screen. So before our editor cuts anything, it watches the footage.

Every clip is logged, shot by shot

When a clip comes in, it is split at its cuts. A long shot is split into steps. One frame from each piece is looked at and described: what it shows, whether it is a photo, a diagram or a graphic, roughly what decade it is from, whether it is sharp, and whether there is a watermark or text on it. Title cards, slates, credits and blank frames are marked as such.

That gives each clip a log, the same thing a human assistant editor would write: 0:00 to 0:06 title card, 0:06 to 0:31 rocket on the pad, 0:31 to 0:40 black, and so on.

The editor places windows, not whole clips

From that log, each clip gets its usable windows: runs of shots that show a real picture, with no title card or black frame in the middle. The editor can only place a clip inside one of those windows. If it asks for a part of the clip that is outside them, the request is refused and it has to pick again.

So the slate does not end up under your narration, and neither does the half second of black at the end of a reel. You do not have to scrub through the render to catch them.

It checks there is enough to cut with

On a film cut from moving footage, before the edit starts, the pipeline adds up how much usable motion each chapter has from those windows. A chapter with plenty of stills and very little moving footage looks like a slideshow. When a chapter has less than 60% of its narration covered by moving footage, the scenes with the biggest gaps go back out for another search, with a description of what they need on screen. Then the editor starts, with a fuller set of footage to choose from.

Scenes that already have a drawn graphic, like a diagram or a map, are left alone. A diagram is the right picture for them, so another clip search would be a waste.

Why this matters for you

Most of what makes AI-assisted video look cheap is not the script or the voice. It is footage that does not match the words, and the small, sloppy things: an end credit flashing up mid-sentence, a black frame, the same shot twice. Logging the footage first is how an editor avoids those, and it is how ours does too.

Cheers, Carl

Next: Picking footage by the facts, not the keywords

Enjoyed this?

This is the tool the post is about. Pick the plan that matches your upload schedule.

See plans and pricing