Craft·June 2026

The First 2 Seconds: How to Engineer Hooks That Stop the Scroll

We once spent close to three weeks on a video that died in two seconds.

Beautiful piece. Shot properly, graded until the skin tones sat exactly where we wanted them, sound designed down to the room tone. The kind of edit where the team gathers around one screen for the final watch and nobody says anything because everyone knows it's good.

We posted it. And the retention graph looked like someone had pushed it off a roof. A cliff at the open, then a long flat line of the few people who stayed — and stayed all the way, by the way, because the video was genuinely good. It just never got the chance to be good in front of anyone.

I've thought about that video a lot since. Not because the loss stung — losses sting for a week and then you move on — but because it forced me to accept something I'd been quietly avoiding: everything we made after second two didn't matter, because almost nobody arrived at second three.

The two-second audition

Here's the frame I now use with every client and every editor on my team: your video is auditioning, and the audition lasts about two seconds.

with a hook you hold; without, it craters.

Pass it, and the platform gives you the next ten seconds. Pass those, and you get the next thirty. It's a ladder, and the first rung is brutally short. Fail it and the insight you spent a week scripting, the b-roll you woke up at 5 a.m. to shoot, the joke that made your team laugh in the edit room — none of it exists. Not "performs badly." Doesn't exist. Unseen work and unmade work look identical from the outside.

Most creators — and I was absolutely one of them — spend nearly all their effort on the part of the video that almost nobody reaches. The body. The middle. The payoff. It feels virtuous, because that's where the substance lives. But it's a strange way to allocate effort once you see the shape of the funnel. The two seconds that decide whether anyone watches get written in five minutes, tired, as an afterthought on the way to export.

At the studio we flip that. The first two seconds get treated as their own production — their own writing pass, their own edit session, their own argument in the review call. This post is basically everything we've learned doing that across well over a thousand videos.

I run a studio that sells editing, so yes, obviously biased about how much the edit matters. But stay with me, because most of what follows costs nothing to apply.

I resisted this for longer than I should have

A confession before the craft talk, because I don't want to pretend I arrived at this cleanly.

I used to think hooks were the tacky part of this business. The clickbait part. I got into this because I love cinema — I wanted to make things that felt like films, and "engineer the first two seconds to stop a scroll" sounded like the opposite of that. It sounded like carnival barking. Early on, when I was undercharging and saying yes to everything, I'd deliver these lovingly cut videos with soft, patient openings, and I'd privately judge the creators who opened every reel shouting a promise at the lens.

Then I watched my patient openings get skipped. Over and over. Not because the audience was shallow — because the context was. Nobody watches a reel the way they watch a film. They watch it mid-scroll, sound off, thumb already moving, standing in a queue somewhere. The two-second audition isn't a moral failing of the audience. It's just the room you're performing in. A theatre audience has paid and sat down. A scrolling audience has done neither.

Once I stopped resenting that and started designing for it, something unlocked that I did not expect: the hook stopped being the enemy of the cinematic work and became its bodyguard. The craft in the body of our videos only gets seen because the first two seconds fight for it. These days I'd say the hook is the most cinematic decision in the whole piece — it's pure film grammar, compressed. What do you show first, and why.

The hook isn't the tax you pay to make good work. It's the thing that lets the good work get seen.

The hook is the first frame, not the first line

The most common mistake I see — in scripts clients send us, in videos I watch late at night when a competitor's reel is bugging me — is treating the hook as a sentence. A line of copy. Something you write.

But people decide to stay before they've heard a full sentence. Often before they've heard a full word. Which means the hook is not the first thing you say. It's the first thing they see.

The first frame, the first movement inside that frame, the first cut — that's the actual hook. The spoken line rides on top of it. When we review a hook at the studio, we look at the opening shot with the audio muted before we ever discuss the writing, because that's the order the viewer experiences it in, whether we like it or not.

What does a strong first frame do? A few things, and it usually only needs one: it shows something mid-action rather than pre-action. It shows a face reacting rather than a face preparing to talk. It has motion — a camera push, a subject moving through frame — because a static frame reads as a pause and a pause reads as permission to leave. Or it shows something slightly wrong, slightly out of place, that the brain wants to resolve.

The talking-head-taking-a-breath opening — you know the one, the creator settling into frame, a half-second of eye contact, then "so today I want to talk about" — that's not a hook. That's a warm-up. And warm-ups get skipped, no matter how good the person warming up is.

So when I say "hook" for the rest of this post, I mean the unit: frame plus motion plus line, landing together in the same beat. Get comfortable with that definition, because everything downstream depends on it.

The three engines

After enough repetitions, we found that nearly every hook that works for us runs on one of three engines. Not formulas — engines. The surface changes every time; the mechanism underneath doesn't.

Curiosity gap. You open a question the viewer can't not want answered. The key word is can't — a good curiosity gap isn't a polite invitation, it's an itch. "3 Dinners in a Day for Weight Loss" is one of ours, and it works because the claim sounds like it shouldn't be true. Three dinners? For weight loss? The brain flags the contradiction and demands the resolution. Note what's doing the work there: not mystery, but tension between two things you already believe. Vague mystery — "you won't believe what happened next" — is the cheap cousin, and audiences have built immunity to it. Specific contradiction still lands.

Stakes plus specificity. You make a promise concrete enough to be checkable and consequential enough to matter. Numbers help. Named difficulty helps. "I tried a hard workout" is nothing; "the world's hardest pushup, in 73 days" is a contract. More on this one in a second, because it deserves its own section.

Pattern interrupt. The scroll runs on autopilot. Thumb moves, eyes half-focus, everything looks like everything else — same framing, same caption font, same beat-matched cuts. A pattern interrupt is anything that breaks the trance: an unexpected location, a frame that's compositionally wrong on purpose, silence where the feed has trained you to expect music, a sentence that starts in the middle. It buys you the shortest window of the three engines, maybe half a second of "wait, what" — but half a second is exactly what you need to deliver the curiosity gap or the stakes right behind it.

That's the honest hierarchy, by the way. Pattern interrupts open the door; the other two engines get people through it. A pattern interrupt with nothing behind it is a doorbell prank.

Seventy-three days, one pushup

Let me take one of our own videos apart properly, because abstractions only get you so far. The piece is embedded below this essay: "Average Indian Learns the World's Hardest Pushup in 73 Days."

That title-and-hook is running stakes plus specificity, and every word in it is load-bearing.

"Average Indian" — that's the identification. Not an athlete, not a calisthenics influencer with four percent body fat. Average. The viewer is being told, in two words, this could be you, which raises the stakes from spectacle to possibility. Watching a superhuman do something superhuman is mildly interesting. Watching someone like you attempt it is personal.

"Learns" — process, not miracle. The word quietly promises a journey with failures in it, which is a more believable promise than "masters" or "achieves."

"the World's Hardest Pushup" — a superlative, but a specific superlative. There is an answer to "which pushup is that?" and the viewer doesn't have it yet. Curiosity gap, smuggled inside the stakes.

"in 73 Days" — and this is my favorite part — not 60, not 90, not "in two months." Seventy-three. A round number reads as marketing; an odd, ugly, specific number reads as reporting. It implies somebody actually counted. The specificity does the trust-building that no amount of sincere delivery could do.

Now stack it: identification, process, superlative, receipt. Four moves in about ten words, and the visual under it is the pushup attempt itself — mid-failure, not mid-setup — so the frame is making the same promise the words are.

Could we have opened that video with a soft montage of training? The early cut sort of did. It was nicer. It was also skippable. The version that went out leads with the promise — and the promise is why anyone stayed to watch the craft.

Why we write the hook last

Here's the part of our process that surprises people the most: we don't write the hook first. We write it last. Almost always.

we build the hook last — from the payoff.

That sounds backwards. The hook plays first; surely you write it first? But think about what a hook actually is: a promise about the best thing in the video. And you cannot promise the best thing in the video until you know what the best thing in the video is — which you usually don't, honestly, until the body is cut.

So the studio workflow runs like this. The editor cuts the body of the piece without agonizing over the opening — a placeholder hook, sometimes literally a slate that says HOOK GOES HERE. Then, once the body is assembled, we sit down for what is internally the most argued-over session of the whole edit: finding the single strongest beat in the footage. Not the three strongest. The one. The moment where, if a stranger walked past the screen, they'd stop.

Sometimes it's the moment we planned for. More often it isn't. It's the accidental line in an interview, the take where something went wrong, the reaction shot nobody scripted. Footage keeps its secrets, and you only find them in the timeline.

Then we reverse-engineer the hook from that beat. What promise does this exact moment fulfill? Write that promise. Find or shoot the frame that makes the same promise visually. That's the hook. Built backwards, from the payoff to the opening, so the two are structurally guaranteed to match.

Hooks written first, before the footage exists, are guesses about a video that hasn't been made yet. Some guesses land. But you're gambling with the ninety frames that decide everything, and I've lost that bet enough times to stop placing it.

You can't write the promise until you know what you're able to keep. Cut the body first. The hook is hiding in it.

A few hooks that died on our desk, since I promised honesty over polish. There was the clever one — a wordplay opening the whole team loved, that we eventually killed because we realized we loved it for reasons that required already knowing what the video was about. A hook that needs context is a hook that doesn't work; the viewer arrives with none. There was the overwritten one, a genuinely great promise that took four seconds to say — right idea, wrong duration, and trimming it hurt more than cutting a scene. And there was the one I fought for personally, an atmospheric cold open I thought was the most beautiful thirty frames we'd cut that month. My editor asked me one question: "what does it promise?" I didn't have an answer. Beautiful and mute. We replaced it, and I sulked for exactly one afternoon.

That question — what does it promise? — is now the first thing said in every hook review we run. If the room can't answer it in one sentence, the hook goes back.

A hook is a promise you start keeping by second five

This is the discipline that separates a hook from clickbait, so I want to be precise about it.

A hook opens tension. The video's job is to resolve that tension — and the resolution has to begin almost immediately. Our internal rule: whatever you promised in second two, the viewer should feel you starting to pay it off by second five. Not fully delivered. Started. Visible progress toward the thing.

Why five? Because the viewer who stayed past the hook made a small bet on you, and a small bet has a short patience window. Open with "the world's hardest pushup" and then spend fifteen seconds on channel intro and backstory, and the viewer feels the gap between promise and delivery widening. They leave — not angry, just done. Worse, the ones who reach the end of a video that never quite delivers learn something about you. They learn your hooks lie a little. And the platforms, in their own mechanical way, learn the same thing: strong opens followed by early collapses get read for what they are.

Overpromising is borrowing attention at an interest rate you can't afford. It works exactly once per viewer.

So hook craft has two halves, and everyone only talks about the first. Half one: make the biggest promise the footage can cash. Half two: cash it fast, and keep making small payments the whole way through. In practice, the seconds right after our hook are usually the second-most-worked part of the edit — proof, motion toward the goal, the first taste of the payoff. The hook gets them in the door. Seconds three through five convince them the room is worth staying in.

The mute test

One tool, stolen from our own review process, that you can use today.

Watch your opening two seconds on mute, with the caption covered. Actually covered — thumb over the text. Now ask: does this communicate anything? A promise, a tension, a reason to exist?

If the answer is no — if what's left is a person in a room starting to talk — you don't have a hook yet. You have a warm-up with good audio.

This test sounds trivial and fails a shocking number of videos, including, for a long time, plenty of ours. It works because it recreates the actual viewing conditions of most of your audience: sound off, attention split, caption unread because reading is effort and the thumb is faster. The muted first frame is the true first impression. Everything else — the voiceover you agonized over, the caption you rewrote five times — is second contact.

When a hook fails the mute test at the studio, the fix is almost never "write a better line." It's visual: start deeper inside the action, choose a frame with the tension already in it, put the evidence on screen before the explanation. The line usually survives the surgery. The frame rarely does.

Isn't this just clickbait?

The objection I'd raise if I were reading this, so let me raise it for you. Doesn't all this engineering reduce content to manipulation? Aren't we just optimizing bait?

My answer, after years of doing this: clickbait is a broken promise, not a strong one. A great hook and a clickbait hook are mechanically identical for two seconds — both open a gap, both make a claim. They separate at second five, when one starts paying and the other starts stalling. Everything in this post assumes the video underneath is worth watching. If it isn't, better hooks will only help more people discover that faster, which is its own kind of justice.

A second objection, subtler: "my audience is different — they're professionals, they're patient, they came for depth." I hear a version of this from almost every coach and founder we start working with, and I understand the instinct, because it's really a hope that their expertise exempts them from the audition. It doesn't. Your existing audience might grant you three seconds instead of two. The audience you don't have yet — the one growth comes from — grants you nothing. Dr. Shilpa Arora had around three thousand followers when we started with her, most of them bots. The million people who follow her now were all, at some point, strangers mid-scroll who owed her nothing. Every one of them was won in a two-second audition somebody engineered.

Third: does this apply to long-form? Mostly yes, with the clock loosened. A YouTube viewer who clicked a thumbnail has already passed one gate, so the opening gets maybe fifteen or thirty seconds instead of two. But the structure is the same — promise, proof beginning, payment schedule. The audition is longer. It's still an audition.

And one place where I'll concede the frame genuinely bends: work that people arrive at deliberately. Our short film BOI BOI doesn't open like a reel, and it shouldn't — festival viewers and reel scrollers are in different rooms. Know which room you're performing in. Most of us, most days, are performing in the feed.

The most leveraged ninety frames in the video

Let me end with the arithmetic that changed how I schedule my own team's time — described in shape, not numbers.

When you improve the body of a video — tighter cuts, better pacing, stronger b-roll — you improve the experience of people who were already watching. Retention lifts a little. Worth doing. We do it obsessively.

When you fix a weak hook, you change how many people ever enter the video at all. You're not improving the experience; you're multiplying the audience that has one. Same footage, same message, same edit — and a different opening can put it in front of a different order of magnitude of people, because everything the platform does downstream keys off those first moments.

Fixing the body of a video nudges retention. Fixing the hook multiplies the audience.

One is addition. The other is multiplication. And yet the default creative process spends weeks on the addition and minutes on the multiplication, which is exactly what we did on that video I opened this essay with — the beautiful one, the three-week one, the one that died on the roof it fell from. The body of that piece was maybe the best work we'd done that quarter. The hook was written on export day, in one pass, by a tired editor, and approved by a tired me.

We recut it, eventually. New opening built backwards from the strongest beat, same body almost untouched. I won't dress up what happened next with numbers I'd be inventing — but the second life of that video is the reason hooks became their own production at With Media, with their own session on the calendar and their own question opening every review: what does it promise?

Two seconds is an insultingly small amount of time to judge three weeks of work. It's also the only judging that happens. I spent years wishing that weren't true. Designing for it turned out to be a much better use of the same energy.

“Average Indian Learns the World's Hardest Pushup in 73 Days” — a hook that front-loads stakes and specificity.