7 Video Hook Ideas for Clips That Stop the Scroll
The fastest way to write a clip hook is to stop inventing one each time and pull from a fixed set of shapes. Seven cover almost everything: contradiction, mid-sentence drop-in, stake-first, named-callout, number-tease, confession, and unfinished-loop. Pick the one that fits your moment, fill in the blank, and you have an opening line in under a minute.
"Make a stronger hook" is the most repeated and least useful advice in short-form. It tells you the goal, not the move. What actually transfers from one clip to the next is a small library of reusable openers, patterns you've seen work, written down as fill-in-the-blanks. Below are seven I reach for, each with the template, a worked podcast example, and the one rule for when to use it.
Why the opening line does most of the work
The first three seconds are where a scroll becomes a watch. Practitioners call the opening three seconds "absolutely critical for social media success" (castmagic), and the widely repeated reason is a Facebook/Meta figure: about 65% of viewers who clear the first three seconds go on to watch at least ten (via EnterpriseTube), directional secondary data, not a platform-audited law, but the direction is not in dispute. A clip that loses the viewer at second one never gets to the line that would have earned the follow.
There's a second reason the opening words matter more than the rest: most social video is watched on mute. Digiday reported around 85% of Facebook video played silently as far back as 2016 (Digiday, publisher-reported and directional), a number that has only become more true. Your first three seconds are usually read off a caption, not heard. So the hook is a writing decision, not a delivery one. You can fix a flat opening at the desk, after the episode is recorded, by changing the line you put on screen.
That's also why clip volume keeps climbing. NPR's "The clipping economy" profiles a paid industry of clippers flooding every feed, with one operator running a network of 40,000 freelance clippers posting on commission. More clips means more competition for that first second, which means the opening shape matters more, not less.
The seven hook formulas
Each formula below has three parts: the pattern (fill in the blanks), a worked example as if it came off a real episode, and the use-it-when rule. Read the moment in your clip, find the matching shape, and write the line.
1. Contradiction, say the thing they don't expect
Pattern: "Everyone tells you [common advice]. That's exactly why you're [bad outcome]."
The viewer's brain flags the mismatch and stays to resolve it. Contradiction works because agreement is boring and disagreement is a question. Worked example, from a fitness show: "Everyone says eat less to lose fat. That's why you keep gaining it back." The clip then earns the claim. Use it when your guest pushes back on conventional wisdom, the contradiction is already in the audio; you're just moving it to the front.
2. Mid-sentence drop-in, start in the middle of the action
Pattern: Open on the line "...and that's when [the bank / my cofounder / the doctor] told me [shocking thing]."
No setup, no "so basically." You drop the viewer into a sentence already in motion, and the missing context is the bait. Worked example, from a business interview: "...and that's when the investor said the company was already dead." The "...and" at the start signals they walked in mid-story, which feels like overhearing something real. Use it when the strongest moment is a confession or reveal buried 30 seconds into an answer, cut the wind-up, start on the verb.
3. Stake-first, name the cost before the content
Pattern: "This [decision / mistake] cost me [specific number / consequence]. Here's what I'd do differently."
People stay for stakes. A vague "let's talk about money mistakes" loses to a named cost. Worked example, from a finance podcast: "This one tax mistake cost me $14,000. Here's the exact form I missed." The number does the work, specificity reads as true. Use it when the clip teaches a lesson; lead with what was lost or won, not the lesson itself.
4. Named-callout, speak to one person, not everyone
Pattern: "If you're a [specific person] who [specific situation], this is for you."
Narrowing the audience widens the watch. A person who recognizes themselves stops; the rest scroll, and that's fine, the algorithm rewards a high finish rate among the right viewers, not a low one among everyone. Worked example, from a parenting show: "If you're a new dad who feels useless in the first month, listen to this." Use it when the advice is for a clear sub-group, name them in the second person so the right viewer feels seen.
5. Number-tease, promise a countable payoff
Pattern: "There are [N] [things]. Most people only know [N-2]. The last one is the one that matters."
A number sets an expectation the brain wants to complete, and "the last one matters" creates a reason to watch to the end, which lifts the finish rate that distribution depends on. Worked example, from a marketing podcast: "There are four ways to price a product. Most founders only use one, and it's the worst one." Use it when your guest lists or ranks things; front-load the count and tease the best item.
6. Confession, admit something costly first
Pattern: "I've never said this publicly, but [admission]."
A confession reads as access, the sense that you're hearing something not meant for an audience. It only works if the admission is genuinely a little risky; a fake confession is obvious and kills trust. Worked example, from a health interview: "I've never told anyone this, but I relapsed twice before this worked." Use it when the audio contains real vulnerability. Don't manufacture it, borrowed honesty is the fastest way to lose the audience you're trying to build.
7. Unfinished-loop, open a question you close at the end
Pattern: "The reason [surprising outcome] happened isn't what you think, I'll get to it, but first..."
You open a loop in second one and close it in the final seconds. The unresolved question is what keeps the thumb still. The risk is the over-tease: if you stall too long or the payoff under-delivers, you train people to distrust your openings. Worked example, from a true-crime show: "He confessed, but not to the crime they arrested him for." Use it when the clip has a genuine twist worth waiting for. The honest version of this is a promise you actually keep.
A decision rule for picking the formula
You don't need to memorize all seven. Read what your strongest moment contains, and the formula picks itself.
The hook is downstream of the moment. If you picked a flat moment, no formula saves it, which is why picking the clip moment that travels comes before writing any opener. And if you want the platform-specific version of the first three seconds, see what to put in the first 3 seconds of a clip.
Common mistakes that kill a good hook
Even with the right formula, four habits quietly waste the opening.
- Throat-clearing. "So, basically, what I wanted to say is..." Cut everything before the first real word. The formula goes in the first frame, not the fourth second.
- A hook the clip doesn't pay off. An over-tease that under-delivers trains viewers to distrust your next opener. If you promise a twist, the twist has to land, virality without a real payoff is empty engagement.
- Vague where specific would land. "A lot of money" loses to "$14,000." "A common mistake" loses to "the tax form on line 12." Specificity reads as true.
- A great spoken hook nobody can read. Most viewers see your opener as a caption on mute. If the text is small, low-contrast, or hidden behind a platform's UI, the hook never registers. Burn it in big and high-contrast.
The fix for the last one is structural, not creative, see why your podcast clips get no views for the distribution and caption issues that sink otherwise-good hooks.
Where AI clip tools fit
A clip tool can find the moment and put a draft caption on screen; it can't yet reliably write the right opener for it. Tools like QuickReel detect strong segments and auto-caption them, and most modern clippers surface roughly the same moments, so the moment-finding is closer to a commodity than the marketing implies (how AI clip detection actually works covers the mechanics). The opening line is still where a human earns the watch. Treat the AI's suggested clip and its auto-title as a first draft, then rewrite the first line using a formula above. The triage habit that makes this fast is in how to pick the best AI-suggested clips.
FAQ
How long should a clip hook be? One line, three to five seconds of speech, or a single readable caption. The hook's job is to buy the next five seconds, not to summarize the clip. If your opener needs two sentences to land, the moment behind it is probably too complex to clip.
Do hook formulas still work if everyone uses them? The shapes don't wear out; lazy execution does. A contradiction or a named cost works because of how attention functions, not because it's novel. What gets stale is the filler version, "you won't believe this" with no actual claim. Keep the structure, make the content specific and true.
Should the hook be spoken by the host or added as text? Both, ideally. Because most clips are watched on mute (Digiday, ~85% of Facebook video silent), the on-screen caption carries the hook for the first second, and the spoken line confirms it for anyone with sound on. Match the two so they reinforce rather than compete.
Can I reuse one hook formula across every clip? You can, but you shouldn't. Reusing one shape makes your feed feel formulaic and trains regular viewers to skip your openers. Rotate through the seven based on what each moment contains, the decision rule above makes that fast.