I timed a conversation last week. Not the episode, the conversation. Two people, 58 minutes, one camera, no music, no cuts to a B roll. It felt quick. Then I opened a different episode from a different show, 41 minutes, better mics, better lighting, three cameras, and it felt like sitting in a waiting room.
The short one felt fast. The long one felt slow. Runtime had nothing to do with it.
That gap is the whole job. Editors talk about pace like it is a number you turn up, as if you can drag a slider marked "energy" and the episode wakes up. It is not a number. Pace is the listener's sense of whether the next thing is worth waiting for. If the answer is yes, they will sit through a nine second pause. If the answer is no, they will bail during a well cut punchline.
Here is what I actually look for after watching a few thousand hours of two people talking into microphones.
Fast is a promise, not a tempo
Every few seconds a listener is running one silent calculation: is something coming? Not "am I entertained right now" but "is this heading somewhere". You can be dull for twenty seconds if the listener believes there is a payoff at the end of the thought. You can be lively for twenty seconds and lose them if it reads as filler.
This is why aggressively cutting an episode sometimes makes it feel slower, which sounds backwards until you have done it. Chop out every breath and every pause and the audio becomes a flat wall. There is no shape left, so there is no way to tell setup from payoff. The listener stops predicting and starts checking the scrubber.
Silence is not dead air. Silence is punctuation. Remove all of it and you have written a paragraph with no full stops.
We do not run silence detection over an episode and let it decide. A human watches the whole thing. That is not a marketing line, it is the only way I know to tell a pause that is thinking from a pause that is nothing, and those two look identical on a waveform.
The first ninety seconds set the clock
Whatever rhythm you establish early becomes the listener's baseline for the rest of the episode. If you open with a cold, tight, specific exchange, a slower stretch at minute twenty reads as "he is building to something". If you open with a slow ramp, a sponsor read, and a round of hellos, the same stretch at minute twenty reads as "this show is always like this".
I have written before about how intros quietly leak listeners, and the mechanism is exactly this. The intro is not just a chunk of time people skip. It teaches them what your show costs to watch. A cheap opening buys you patience later.
Practical version of that: find the moment in the first ten minutes where the guest says something the host clearly did not expect. Put it at the top. Do not explain it. Do not caption it "coming up". Just start there and let the episode catch up.
Overlap is the sound of a real conversation
The most common thing that makes an interview feel slow is not the talking. It is the handoffs.
Two people who are being polite leave a beat between turns. Every turn. Multiply a half second of politeness by four hundred turns and you have added three minutes of nothing, spread so thin that nobody can point at where it went. The episode does not feel long in any one spot. It feels long everywhere.
Tightening those handoffs is the single highest yield edit in podcast work. On a two person show I will:
- Trim the gap between turns to roughly the length of a natural breath, not a natural pause.
- Let the incoming voice start slightly under the outgoing one where the tone allows it, so the turn overlaps by a few frames.
- Keep the longer gaps deliberately, but only where the previous line earned a moment.
- Leave laughter alone. Laughter cut short is the most obviously fake sound in editing.
Do this well and hosts tell us the episode feels shorter without anyone being able to say what changed. That is the goal. Good pacing is invisible. If a listener notices the editing, something went wrong or something went very right, and it is usually the first one.
Cameras change rhythm even when nobody is moving
On a multi-cam episode you have a second lever that single camera shows do not have. A cut to the listening face is a beat of time that costs nothing in runtime but resets attention. Used sparingly it makes a long answer feel like a scene instead of a monologue.
Used badly it does the opposite. Cutting on a metronome, every four to six seconds regardless of what is being said, is the fastest way to make an episode feel like a corporate training video. The camera should move when the meaning moves. A cut in the middle of a thought splits the thought. A cut at the end of one, right as the other person reacts, makes the whole exchange land.
Single camera shows are not stuck, by the way. Punch ins, a held wide, and the willingness to sit still are enough. Some of the fastest feeling episodes I have cut were one static camera and a host who knew when to shut up.
The graph tells you where, not why
Retention graphs are the honest part of this. They will show you the exact second people leave. What they will not tell you is the reason, and this is where most people get it wrong: the drop is almost never at the boring part, it is a few seconds after the moment that made the boring part feel pointless.
I go into the detail in how to actually read a retention graph, but the short version for pacing is to scrub backwards from the cliff, not forwards. Find the promise the show made, then find the moment it quietly dropped that promise. That gap is your edit.
One more thing about the graph. A flat line with small dips is a better sign than a high average with one big cliff. Flat means the rhythm holds. A cliff means one specific betrayal.
What breaks the feeling instantly
Some things kill pace no matter how good the edit is, because they break the listener's trust in the room:
- Audio that drifts out of sync with the picture. Even a few frames reads as "wrong" before anyone can name it. If you have ever fought this, the drift fix workflow is worth keeping around.
- Wildly different levels between two speakers, so the listener adjusts volume twice a minute.
- A transcript based cut that removed half a word because the machine got the boundary wrong. This is a real risk and the reason I still do not hand the timeline to AI transcription alone. Text accuracy and cut accuracy are different problems.
- Music that starts under a sentence and ends nowhere.
None of these are pacing problems on paper. All of them make an episode feel slow, because the listener spends attention on the mechanics instead of the conversation.
Clips are a rhythm test you can run for free
If you want to know whether a conversation has real momentum, try pulling a 45 second clip for LinkedIn out of the middle of it. Not the best moment. The middle.
If you can find a self contained thought with a setup, a turn, and a landing inside any random two minute window, the episode is paced well. If every candidate clip needs thirty seconds of context glued to the front, the episode is not slow because of the edit. It is slow because the conversation kept starting over.
That is a hosting note, not an editing note, and it is the one thing an editor cannot fully rescue. The fix happens in the room: finish the thought, then move. Do not restate the question before answering it. Do not summarise what you just said.
What I would change first on most shows
If a host asked me for one change, in order, this is the list.
- Cut the first two minutes and start on the surprise.
- Tighten every handoff by a few frames.
- Keep three long pauses on purpose and remove the rest.
- Stop cutting cameras on a timer.
- Fix levels before you fix anything creative.
Nothing on that list is expensive. Nothing on it requires new gear. All of it is attention, which is exactly why it is the part worth paying a human to do.
A conversation feels fast when the listener trusts that the next sixty seconds will be worth more than the last sixty. Everything else is craft in service of that one bet. If you want to know how we work through this on an actual episode, the process is written out plainly, and if you just want a second opinion on a cut you already have, send it over.