I spent a Tuesday morning last month watching an editor tighten a 62 minute interview down to 51. Good work, clean cuts, nothing obviously wrong. Then the host wrote back and said the episode felt like it was chasing him. He could not point at a single cut. He just did not want to listen to it twice.
We went back and put nine pauses in. Not new material. Just breath that had already been recorded and then removed. The episode got longer by about forty seconds total, and it stopped feeling like a treadmill.
Silence is content. It is the only piece of content most editors are trained to delete on sight.
The pause is where the listener catches up
Speech carries more information per second than most people can absorb in real time. That is fine at first, because you have runway. Ten minutes into a dense conversation, the listener is behind. They are still turning over the thing the guest said about pricing while the host has already moved to hiring.
A pause is the moment they catch up. Half a second after a real idea lands, the brain does something with it. Cut that half second and the idea goes past unprocessed. The listener does not think "that was too fast". They think "I am tired" and they leave.
That is the retention story nobody puts in a thumbnail. Most drop-off is not caused by a bad question. It is caused by the accumulated cost of never being allowed to stop.
What automation gets wrong about silence
The reason pauses vanish is that removing them is trivially easy to automate and impossible to automate well. A waveform tool sees amplitude below a threshold for longer than a set duration and cuts. It cannot tell the difference between a guest thinking and a guest waiting for the barista to stop grinding beans.
We do not run silence detection on anything. Human editors watch the whole episode, start to finish, before a single cut is made. That sounds inefficient and it is, slightly. It is also the only way anyone can tell that the 1.2 seconds at 14:07 is dead air and the 0.9 seconds at 22:41 is a guest deciding whether to say the true version of the answer.
The second one is the best moment in the episode. An algorithm will take it out every time, because to a waveform they are identical.
A pause before an honest answer is not empty. It is the sound of someone deciding to tell you something.
I have made this mistake myself, early on, in the direction of speed. If you are curious how much of the finished feel comes from choices like this, I wrote about the editing decisions nobody notices, because pause handling sits right at the top of that list.
The four pauses worth protecting
Not all silence earns its place. Most of it genuinely should go. Here is the split we work with internally.
- The thinking pause. Before an answer, when the guest is actually composing something. Keep almost all of it. Trim the tail if it runs past about two seconds.
- The landing pause. After a strong line. This is the one editors kill most often, because the line is over and the instinct is to move. Keep half a second to a full second. It tells the listener that what they just heard mattered.
- The turn pause. Between topics. Acts like a paragraph break. Without it, an hour of conversation becomes one undifferentiated block.
- The reaction pause. The beat before laughter, or the beat where the host clearly has to physically recover. Related problem, different craft, which is why editing laughter without flattening it is its own conversation.
Everything else goes. Throat clearing, the gap while someone finds a tab, the four seconds after a question that landed badly and needs rescuing. Those are not pauses, they are holes.
Video makes this harder, not easier
In audio, a pause is just quiet. In video, a pause is a face. That changes the maths in both directions.
On camera, a two second thinking pause with a good reaction shot is watchable, sometimes the most watchable part of the episode. The same two seconds on a locked wide shot where nobody moves is unbearable. This is where multi-cam editing earns its money: the second angle gives the pause somewhere to live. You cut to the listening face, and silence becomes attention instead of absence.
With one camera you have fewer moves, so the trims get tighter and more surgical. It is doable. It just requires more discipline, not less, because you cannot hide behind an angle change.
The platform matters too. A pause that reads as thoughtful on a podcast feed can read as a stall on a video feed where the next thumbnail is one flick away, which is part of why the YouTube versus Spotify decision changes how aggressively we cut the same source footage.
What we actually do on a first pass
Rough order, and this is the part people ask about when they want to know why a properly edited episode takes the time it takes.
- Watch the whole thing. No timeline, no cutting. Just notes on where it drags and where it hums.
- Remove the holes. Restarts, dead air, technical fumbles, the guest apologising for their dog.
- Cut for structure. Whole tangents in or out. This is the biggest single lever on retention and it has nothing to do with silence.
- Pass over the pauses one at a time and ask what each one is doing. Keep, trim, or kill.
- Watch again at normal speed. Not at 1.5x. If it feels breathless, we put pauses back.
Step five is the one that gets skipped everywhere. Editors watch their own work sped up because they have already seen it four times, and at 1.5x every pause feels too long. You end up cutting an episode for an audience of one person who has heard it all before.
Our first cut lands in 48 hours, revisions come back the next working day, and pacing notes are the single most common revision request we get. Not colour, not audio. Pacing. Usually the note is some version of "it feels rushed", and the fix is almost always adding time back rather than taking more out.
Clips are the exception, and that is fine
I am not arguing for slack everywhere. A 40 second vertical clip has no room for a landing pause, because the viewer arrived by accident and is deciding in under two seconds whether to stay. In short clips we cut breath to the frame, because the job is completely different: the clip has to survive without context.
That is the point. Pacing is not a setting you dial in once. It is a function of where the thing is going to be watched. The full episode is for someone who chose to be there. The clip is for someone who has not chosen anything yet. Cutting both to the same rhythm is how shows end up with episodes that feel like ninety minutes of advertising for themselves.
Packaging follows the same logic, incidentally. The thumbnail and title are doing the accidental-viewer job, which is why our sister studio ClickTheory handles that side separately rather than treating it as an editing afterthought.
The show that needs pauses most is the one that fears them most
New hosts cut hard. I understand why. You listen back to your own voice, you hear every gap as a failure, and you compress it into something that sounds professionally busy. Then you wonder why nobody finishes.
Established hosts do the opposite. Listen to any interview show that has been running for six years. The pauses are enormous. The host will sit in silence for three full seconds after a guest says something heavy, and it is riveting, because the silence signals that the host is not in a hurry, which signals that what was just said deserves the time.
Confidence sounds slow. That is not a style choice, it is a signal, and audiences read it accurately.
If you are running more than one format, the difference gets sharper still. A tight solo show and a long interview want opposite pacing, which is one of the things to work out before you go splitting your show into a second channel and end up cutting both the same way out of habit.
How to test this on your own show
You do not need us to check this. Take your last episode and do three things.
- Find the best moment. Check what happens in the second after it. If the next word starts immediately, you cut too close.
- Look at your retention graph at the point where it starts sagging. Listen to the two minutes before it, not the moment itself. Drop-off is a lagging indicator of pacing.
- Play thirty seconds at normal speed with your eyes closed. If you feel like you are being chased, your listeners feel it too.
Then put four pauses back and see if you can hear the difference. Most people can, immediately, and cannot un-hear it afterwards.
I am not romantic about this. Plenty of silence is just waste and should be cut without ceremony. But the default setting across podcast editing right now is set way too aggressive, mostly because the tools made deleting silence free and nobody stopped to ask what it cost.
If your show feels fast and you are not sure why, that is usually where I would look first. Happy to listen to an episode and tell you what I hear.