How to avoid dead air

Dead air is the gap between your guest finishing and you starting. Two seconds of it feels like ten in the room, and on camera it reads like ten too. Everyone who records interviews hits it. Most of the advice about it amounts to "be more prepared," which is not advice.

So here is what actually causes it, and what to do in the moment. None of the techniques below need software, and they work on your next recording. The last section is about the one kind of dead air technique cannot fix, which is the part we build for.

Why you blank

Three different things get called dead air, and they have three different fixes. Knowing which one just happened is most of the work.

You are doing two jobs at once

Hosting means listening for meaning and planning your next move at the same time. Both run on the same attention. When a guest says something genuinely surprising, listening takes all of it, and when they stop, the planning half has nothing queued. The better the answer, the more likely you blank after it. That is a compliment you cannot use.

The guest answered in eight words

You budgeted three minutes for that question and got a sentence. Now you are holding a gap you did not schedule. This one is not a memory failure at all. It is a supply problem, and it needs a move you can run from any answer, not more preparation.

You lost one specific thing

A name, a year, a number, the title of the episode where you covered it. You have plenty to say. You just cannot retrieve the single item the sentence needed. Nothing about breathing, rapport, or note-taking discipline fixes this in the two seconds you have.

The follow-up ladder

This is the move for the eight-word answer. Three rungs, in order, and you can climb them from literally anything a guest says. Memorize the ladder instead of a list of questions.

Feeling

"What was that like at the time?" You move them from what happened to what it was like to be there. Almost nobody answers this one briefly, because it requires them to go back and look.

Example

"Give me a specific instance." Abstractions are short. Stories are long. If a guest says the market got tough, ask for the week it got tough, and you get a scene instead of a summary.

Contrast

"Who disagrees with you about that, and what is their best argument?" Or: "What would have had to be true for the opposite to happen?" This is the rung that produces the clip. It is also the rung most hosts skip because it feels confrontational. Asked with real curiosity, it never lands that way.

Because each rung is built out of the answer you just heard, none of them sound like you are reading from a page. Ask one at a time. Two questions stacked together lets a guest pick the easy one and hand you back another short answer.

The bookmark phrase

A bookmark phrase buys you three to five seconds honestly, by handing the floor back to the guest instead of stalling in front of them.

The reliable version is repeating their own last noun. "Say more about the funding round." "Back up to the part where you quit." It costs you no invention, it is unmistakably active listening, and it ends on them talking rather than on you searching.

The other version parks something you want to return to. "Hold that thought, I want to come back to what you said about the second factory." Now you have a thread you can pull whenever the next gap opens, and it is a thread the audience already heard, so returning to it feels planned.

What does not work: "that is so interesting," "wow," "amazing." They buy about one second, they end with you still holding the floor, and audiences hear them as the sound a host makes when they were not listening.

Pre-load three escape hatches per segment

Before you record, take each segment of your rundown and write three things you could go to cold, with no setup: a question you would be glad to ask out of nowhere, one specific fact or number, and one callback to your own past work.

For a typical episode that is nine to twelve items, and they fit on one page you can keep in view. You will use two of them. That is fine and expected. The value is not the items, it is that you are never at zero, and knowing you are never at zero is what lets you take a risk on a follow-up instead of retreating to your script.

Write them as fragments, not sentences. A fragment you glance at gets rebuilt in your own voice. A full sentence you glance at gets read aloud, and everyone can hear the difference.

Silence that is fine, and silence that is not

Not every gap is a problem. After you ask a hard question, count to three before you rescue anybody. New hosts fill that space almost every time, and it costs them the best material, because the real answer is usually the second sentence, the one that arrives after the guest actually thinks. The beat after a heavy answer is worth protecting too. Stepping on it makes you look like you were waiting rather than listening.

The bad kind is when the guest is waiting on you and cannot tell whether their answer landed. That is the silence that makes people fill space with a worse version of what they already said. The other bad kind is silence with your own filler rising through it, the "um, so, yeah, right" that tells everyone you are searching.

One test settles it. Ask who the silence belongs to. If it belongs to the guest thinking, leave it alone. If it belongs to you looking for something, close it.

Editing does not solve this

If you publish audio only and cut heavily, you can remove the gap. Fewer and fewer shows are in that position. Video shows the jump, live streams have no post, and clips get pulled from the raw take by people who do not care what your edit looked like.

The deeper problem is that the two seconds are not the real cost. What it does to the next five minutes is. After you blank, you get careful. You stop reaching for the follow-up you are not certain you can land, you steer back toward your prepared list, and the interview flattens out. Cutting the gap in post does not recover the question you were too rattled to ask.

The name you lost

Everything above handles the first two causes. It does nothing for the third. When the missing piece is a name, a date, or the episode number where you said the opposite thing, you do not need a better question. You need a lookup, and there has never been one available while your mouth is moving.

Broadcast solved that a long time ago and did not call it a technique. The anchor has a control room, and someone in it feeds them the name before the pause is audible.

That is the problem Tally exists for. Live intelligence means the right information appearing on screen while you are still talking. The desktop app listens to your session and puts a short line of text in front of you: the guest bio when you fumble their name, the episode where you covered a topic before, a flag when a number that just went by is wrong. It makes no sound, it stays out of your recording chain, and you never break eye contact to go looking.

It runs on macOS today, and it is invite only while we onboard shows one at a time. Here is how live intelligence works.

---

Tally is invite only right now

We are onboarding shows in small waves so we can set each one up properly. Leave your email and we will reach out when the next wave opens.