How long should a pause be? Notes from 400 talking-head videos
We measured the gaps between sentences in videos that hold attention. The median sits around 0.35 seconds, and the spread tells you more than the median.
Every editor has a number in their head for how long a pause should be. Almost nobody has measured it. We did, on a sample of 400 short talking-head videos from creators whose retention we could see, and the result is less a number than a shape.
What we measured
We transcribed each video with word timestamps and looked at the silence between the end of one sentence and the start of the next. Breaths, filler words and restarts were counted as part of the gap, because that is how a viewer hears them. We ignored gaps that overlapped a visual change, since a b-roll shot changes what a pause means.
That leaves about 31,000 gaps. Here is where they fall.
| Gap between sentences | Share of gaps | What it sounds like |
|---|---|---|
| under 0.2 s | 18 % | breathless, cut-on-every-word |
| 0.2 – 0.5 s | 47 % | conversational, edited |
| 0.5 – 0.9 s | 24 % | deliberate, teaching |
| 0.9 – 1.5 s | 8 % | dramatic, or a lost thought |
| over 1.5 s | 3 % | dead air |
The median is 0.35 seconds. The videos with the flattest retention curves cluster in the 0.2 to 0.5 band and spend their long pauses on purpose.
The spread matters more than the median
A video where every gap is 0.35 seconds sounds like a machine. The videos that held attention had a tight centre and a few long outliers, and the outliers were in predictable places: before a number, after a question, before the answer to that question.
That pattern is what a good editor produces by hand, one pause at a time. It is also exactly what a pause dial with pinning produces: set the target, then pin the four or five pauses that should stay long. The dial takes the centre; you keep the outliers.
A rule you can use
- Default every pause to 0.4 seconds. It is inside the conversational band and leaves room for a breath.
- Pin the pause before your most important sentence. Let it run to a full second.
- Kill everything over 1.5 seconds unless something is happening on screen. That is the dead-air line, and viewers agree on it with remarkable consistency.
- Never go under 0.2 seconds between sentences unless the two sentences are one thought. Below that the cut is audible as a click and the speaker stops sounding like a person.
Why this is easier in text
On a waveform, a pause is a flat stretch you have to find by eye and trim by hand, twice per gap. In a transcript, the pauses are already there between the words, with their lengths attached. Changing all of them is one gesture. Changing one of them is a tap. That is the difference between knowing the right number and being able to apply it to a video before you lose interest in the video.