Last month I timed the cold opens on eleven episodes we were editing in the same week. The shortest was 6 seconds. The longest was 2 minutes and 41 seconds. The 2:41 one belonged to a show whose host told me, in the same message, that people kept dropping off before the intro finished. I did not need analytics to guess why.
Here is my position, and I will defend it for the next 1,700 words: a cold open should be between 12 and 35 seconds, and it should be a single unbroken moment, not a montage. Everything outside that range needs a reason. Most shows do not have one.
What a cold open is actually for
A cold open is not a trailer. It is not a highlight reel. It is not a place to prove the episode was worth recording.
A cold open does one job. It makes the listener decide to stay through the intro music. That is the whole assignment. You are not selling the episode, you are buying yourself the next 30 seconds.
Once I started thinking of it that way, the 2 minute cold opens stopped making sense to me. If you need 2 minutes to convince someone to stay for a 45 second intro, the intro is the problem, not the cold open.
The mistake I see most is treating the cold open as a summary. Hosts pull the four best lines from across the episode, stack them, and call it a hook. It sounds impressive in the edit bay. In a car, on a phone, at 8 in the morning, it sounds like four strangers talking at you.
A cold open made of four great moments is worse than a cold open made of one good one. Compression is not the same as intensity.
The 12 to 35 second range, and why
Under 12 seconds you almost never get a complete thought. You get a fragment that sounds punchy and means nothing. "And that is when I realised the whole thing was a lie." Great. About what? The listener has no anchor, so the line lands as noise. Fragments work in trailers because trailers are three minutes of fragments and the form is understood. One orphan fragment at the top of a podcast just reads as a mistake.
Over 35 seconds you stop being a hook and start being content. And content at the top of an episode, before any orientation, is hard work for the listener. They do not know who is speaking. They do not know the stakes. They are doing comprehension labour before they have agreed to.
Inside 12 to 35 seconds you can fit exactly one thing: a setup and a turn. Somebody says what they believed, and then says what actually happened. That structure is old and it works and it fits in about 20 seconds of natural speech.
When I am picking a cold open I look for:
- A complete sentence with a subject the listener can identify without context
- A turn, a reversal, a number, or an admission
- Clean audio with no cross talk, because the first 10 seconds is where people judge your production
- Energy in the voice that is a little above the room average for that episode
That last one matters more than people expect. If your cold open is the calmest 20 seconds in the episode, the episode sounds like it is going to be slow even when it is not.
Where the good ones actually live
Not the beginning. Almost never the beginning.
The first 10 minutes of most recordings are people finding their footing. The energy is polite. The answers are rehearsed. I have written before about why guests sound better on the second take, and the same physics apply here: the guest has to forget the microphone before they say anything you would want to open with.
The good cold opens sit between minute 20 and minute 50 in a 60 minute conversation. Deep enough that the guest has stopped performing, early enough that they have not started summarising themselves.
There is a second reliable spot: right after a laugh. People say true things in the two seconds after they stop laughing, because the guard is down. Some of the best cold opens I have cut were the sentence immediately following a laugh that I then trimmed off the front.
This is also the exact reason I do not believe in machine-selected cold opens. A tool that ranks moments by volume, waveform energy, or keyword density will hand you the loudest 20 seconds, which is usually somebody making a joke about coffee. It will not hand you the quiet admission at minute 38. That distinction is most of the job, and it is a big part of why auto-cut podcasts feel wrong even when nothing is technically broken. Our editors watch the whole episode, start to finish, before anybody picks a cold open. There is no shortcut that survives contact with a real conversation.
The four cold opens that keep showing up
After a few hundred episodes you start seeing the same shapes. These are the four that work.
The admission. The guest says something slightly against their own interest. "We burned through the first 40 grand in four months and I had no idea what I was doing." Instant credibility, instant curiosity. My favourite, and the hardest to find, because it only happens once the guest trusts the room.
The number. A specific figure with a claim attached. Not "we grew a lot", but "we went from 11 clients to 90 in one year without hiring anyone". Numbers work because they promise the episode has substance rather than vibes.
The disagreement. The host pushes back and the guest does not fold. Twenty seconds of two people actually not agreeing beats two minutes of two people being nice. This one only exists if the host is willing to push, which is a recording problem, not an editing problem.
The scene. A short piece of storytelling with a physical location in it. "It is 2am, I am sitting in a car park in Sharjah, and my phone will not stop ringing." Scenes work on audio the way nothing else does, because the listener builds the picture and then wants to see how it resolves.
What is not on that list: the guest introducing themselves, the host explaining what the show is about, and anything that begins with "so".
The cut itself, and the five seconds after it
Picking the moment is half of it. The other half is what happens when the cold open ends.
Three rules I hold to.
Do not fade the cold open out. A fade tells the listener the moment was not important enough to finish. Cut it clean on the last consonant, leave about 300 milliseconds of room tone, then hit the intro.
Do not put music under the cold open on the first episode of a show. A bed under raw speech is a production choice that reads as confidence when you have earned it and as compensation when you have not. Let the words be dry. Add the bed later once the show has an identity, at which point it feels like a signature. If you are building that identity from scratch, that is a conversation about intros and branding rather than a cold open decision.
Do not let the intro immediately repeat what the cold open just said. It is the single most common flow break I fix. The guest delivers a great line about losing 40 grand, and then the host says "today we are talking about the mistakes founders make with money". You just explained the joke.
The handoff should add, not restate. Cold open gives the moment, intro gives the context the moment was missing. If your intro works as a standalone paragraph after any cold open, it is generic and it is costing you.
Video changes the maths, but less than you think
On YouTube the cold open competes with a thumbnail and a title that have already done most of the persuading. Somebody clicked. They have committed something. So the video cold open can be slightly shorter, because you do not need to win as hard.
What video does demand is a visual reason to stay in those first seconds. A locked wide shot of two people at a table is the least interesting frame in the world. On multi-cam episodes I want at least one angle change inside the cold open, because motion holds the eye while the words do the actual work. On single-camera shows I get that from a push in, a b-roll insert, or a punched in reframe.
Worth saying plainly: on YouTube, the packaging outranks the cold open. If your click through rate is under 4 percent, no cold open on earth will fix your numbers, and you should be fixing the thumbnail first. That is why we send that work to our sister studio ClickTheory rather than pretending an edit can solve a packaging problem.
Make it a rule so you stop deciding
The hardest part of cold opens is not craft. It is that they are a fresh judgment call every single week, and fresh judgment calls are what kill shows around episode 14.
So turn it into a standing rule. Ours, for most shows we cut, is written down in about four lines: 12 to 35 seconds, one continuous moment, sourced from minute 20 onward, hard cut into the intro, no restatement. The editor applies it without asking. The host stops relitigating it.
Rules like this are the difference between a show you run and a show that runs you, which is really the whole argument in building a show that survives your schedule. Fewer decisions per episode, more episodes.
If you want the rule applied consistently by somebody who is not you, that is the shape of our podcast editing packages: a human watches the whole episode, picks the moment, and delivers a first cut in 48 hours. You give notes, we turn revisions around the next working day.
What I would do this week
If you have published more than ten episodes, go listen to your last five cold opens back to back with a stopwatch.
Ask three questions. How long is each one. Could a stranger understand it without any context. Does the intro immediately repeat it.
My guess, and I have made this guess a lot, is that at least three of them are too long, at least one is a montage, and at least two get restated by the intro within ten seconds. Fixing those three things costs you nothing in production time and changes the first impression of every episode you release from here.
The cold open is 20 seconds of a 45 minute episode. It is also the only 20 seconds most people will ever hear. Spend the attention there.
If you want a second opinion on yours, send me an episode at samy@pixelsamy.com and I will tell you where I would have cut it.