12 Viral Clip Hook Formulas That Stop the Scroll Every Time

The decision happens before your first sentence ends

Most viewers decide in under two seconds. Not consciously — it's closer to a reflex. And the thing that survives that reflex is almost never a good clip that starts slowly. It's a clip whose first frame creates an unresolved state in the viewer's head.

Which is why the most common clipping mistake is starting a clip where the interesting part starts. The interesting part is usually the answer, and answers don't hook. Questions hook. Start of a sentence that promises a conclusion hooks. A number with no context hooks.

The twelve formulas below are patterns for creating that unresolved state. They're not templates to write — you're not scripting anything, you're finding moments in existing footage where a speaker already produced one of these shapes. Your job is recognition, not invention.

Each one below comes with the way it fails, because every hook pattern has a version that annoys people instead of holding them.

Formulas 1–6: curiosity, contradiction, and stakes

1. The interrupted claim. Cut in mid-sentence, right before the conclusion. "…and that's the reason most people never—". The viewer has to stay for the ending of a sentence they've already started processing. *Fails when* the conclusion isn't worth the wait, which trains your audience to distrust your clips.

2. The number with no context. "Four hundred thousand dollars, gone in a weekend." A specific figure with no frame demands a frame. *Fails when* the number is vague or rounded — "a lot of money" does nothing.

3. The direct contradiction. Speaker disagrees with something the viewer believes is settled. "Compound interest is not the reason he's rich." *Fails when* the contradiction is manufactured and the clip walks it back immediately.

4. The name drop. The clip opens with a recognizable person or brand being discussed. Works because recognition buys attention before comprehension. *Fails when* the name is bait and the clip isn't really about them.

5. The confession. "I've never told anyone this." Highest-performing opener in the interview format, because it signals scarcity. *Fails when* what follows is mundane. This one burns audience trust faster than any other.

6. The stakes reveal. Open on the consequence, not the setup. "If that call had gone the other way, the company was finished." *Fails when* it's the only thing in the clip and the payoff never arrives.

Notice that four of the six fail the same way: the hook writes a check the clip doesn't cash. That's the actual craft here — matching hook intensity to payoff intensity. Why three-second hooks are overrated makes the case that the payoff matters more than most clippers think.

Formulas 7–12: reaction, information, and format

7. The visible reaction. Open on someone's face doing something — laughing, recoiling, going silent. Faces are processed faster than speech, so this hooks before a single word registers. Especially strong in podcast and interview clips.

8. The list promise. "There are three ways this goes wrong and everyone only knows one." Sets an explicit completion condition, which is why list clips get saved and rewatched.

9. The mid-argument entry. Two people already disagreeing when the clip starts. Conflict in frame one, no setup required. *Fails when* it's manufactured drama; audiences detect that quickly.

10. The correction. "Everyone quotes that study wrong." Positions the viewer as about to become the person who knows better. Strong sharing driver.

11. The visual anomaly. Something in frame that shouldn't be there. Works entirely without dialogue, which makes it the only hook on this list that survives silent autoplay with no captions.

12. The unfinished demonstration. Someone is halfway through doing something. The viewer stays for the result. Native to cooking, fitness, gaming, and any process-based content.

Most strong clips use two of these at once — a visible reaction plus an interrupted claim, or a number plus a contradiction. Stacking three usually reads as trying too hard.

Finding these in an hour of footage

Recognition is easy. Scanning ninety minutes of podcast to find six of these is not — that's the part that eats a clipper's week.

Two workable approaches:

Manual, with the transcript. Pull the transcript, read it rather than watching, and mark every place a speaker produces one of the twelve shapes. Reading is roughly six times faster than watching. Then jump to those timestamps and check whether the delivery matches the text — plenty of great sentences are delivered flatly and don't work as clips.

Automated. AutoClip scans the whole source and surfaces the moments most likely to hook a viewer, with a virality score and a transparent five-criterion breakdown so you can see *why* a moment was ranked highly rather than trusting a number. A typical video comes back in about 10 to 15 minutes with around nine candidates; multi-hour sources take proportionally longer.

Either way, the same rule applies afterward: adjust the in-point. The most common single improvement to a clip is moving the start two or three seconds later, because almost every moment has a run-up that isn't doing anything. Trim to the first word that creates the unresolved state.

The timeline editor exists for exactly this — nudging the entry point is a small edit with a large effect on retention.

Testing hooks without guessing

Take one strong moment and export it three times with three different in-points. Post them across a two-week window on the same platform at the same cadence. The differences in first-two-second retention will be obvious and larger than you expect — often 20 or 30 percentage points from a two-second change.

Do that four or five times and you'll have a personal model of what your specific audience responds to, which beats any general list including this one. Audiences differ. A finance audience responds to numbers and contradictions; a comedy audience responds to reactions and mid-argument entries.

What not to do: change the hook and the caption and the length and the posting time in the same test. You'll learn nothing. One variable.

And track the right metric. Not views — first-two-second retention, or whatever your platform's equivalent is. Views are a downstream consequence of a dozen things. Early retention is the thing the hook directly controls, and it's the number that tells you whether the change worked.