You leave a take in when the stumbles carry more truth than a clean retake would. Editors cut toward perfection by default, and perfection is often the least interesting version of a person talking.
The skill is not cutting. It is deciding which rough edges are load-bearing and which are just drag.
A second pass fixes the grammar and kills the energy. The first time you said the line you meant it. The fifth time you recited it, and viewers hear the difference even when they cannot explain it.
Watch for the take where the voice flattens. Often the messy earlier attempt is the one that moves people, and the perfect one is the one they skip past.
A pause, a small laugh, or a caught breath can stay if the point still lands. A lost thread or a repeated word stops the viewer, and that kind of stumble has to go.
The test is comprehension, not cleanliness. If the mess does not cost the viewer the idea, it might be worth keeping.
The goal of the edit is a person, not a transcript. Some slips are the person showing through.
Watch the take once with the sound on and once with it muted. If the body language in the messier version is livelier, keep it and fix the audio around it.
When you cannot decide, show both to someone who has not heard either. Their reaction to the rough one will tell you more than another hour of your own judgment.
Jump cuts used to read as amateur mistakes. Now they carry most online video, and the reason is less about taste than about how people watch.
Hiding a jump cut with b-roll is the oldest trick in the book. It only works when the b-roll has a reason to be there beyond covering a seam.
Jump cuts for energy tighten a slow take into something worth watching. Push the same trick too far and the story stops being a story at all.
Analyze a channel, extract its structural DNA, and generate scripts engineered from patterns that actually work.
Analyze a Channel