The caption is permission to pause. Not permission to watch.
A thumb moves through a feed fast. The caption doesn't register. The motion does. The colour does. The shape does. The caption is the thing you read after you've already decided to stop.
Every business I work with says the same thing: “The caption will explain it.” No. The caption will add to something that already moved you. It will not save a dead first frame.
What works
Motion. A hand entering from the left. A lid lifting. A tool dropping. The eye tracks movement before the brain decides anything. Start with a static logo and you're asking people to read and decide you're worth watching. Start with motion and you're showing them.
Conflict. Before and after. Broken and fixed. Wrong and right. We stop for contrast. “Here's how it was” — and then it changes. That change is the hook.
Speed. Something already in progress. Not the setup — the action. Not “let me show you how to make this” — the knife already moving. The viewer steps into a story that's already started.
Curiosity. Not answered in the first frame. If frame one shows the whole answer, there's no reason to watch frame two.
What doesn't work
A logo. A face saying hello. A title card explaining what's about to happen. A product shot waiting for context. These are captions pretending to be videos.
The caption that says “wait for the end” is the caption admitting the first two seconds are dead.
Hooks you can steal
- Start at the middle. Cut the first five seconds of whatever you filmed. The video usually starts there.
- Show the result, then rewind. One second of the finished thing, then “here's how”.
- Name the mistake. “Everyone does this wrong” — shown, not just said.
- Put the problem in frame. The leak, the mess, the empty shelf, already on screen.
- Say the specific thing. “This took nine hours” beats “this took ages”.
What a hook looks like in a real business
- A bakery: not the shop sign — a tray sliding out of the oven, already golden.
- A glazier: not the van — the crack across a pane, then the new glass going in.
- An accountant: not a desk — a carrier bag of receipts tipped onto the table.
- A café: not the menu — milk being poured into the espresso, the pattern forming.
- A holiday cottage: not the front elevation — the door swinging open onto a lit fire.
Every one of these is something the business already does. The hook isn't a trick you add; it's choosing the most alive second of what you already film and putting it first.
Sound is part of the hook
For people watching with sound on, the first noise counts as much as the first image: a sizzle, a snap, a pour, a door, a voice mid-sentence. Silence or a slow music intro gives the thumb a reason to move on. Record the real sound of the work and let it play from frame one. Then add captions, so the video still works for everyone watching muted.
How to test your opening
Pause your video on the very first frame. Ask: would a stranger know something is about to happen? Then watch the first two seconds with the sound off. If it still pulls you in, you've got a hook. If not, trim until it does. Most openings improve simply by starting later.
Then check your numbers. If a large share of viewers leave in the first two seconds, the hook failed — no matter how good the rest was. Our guide on where to start covers what else to measure.
Why this matters
A platform is not a gallery. Reels, TikTok and Shorts are feeds, not collections. The thumb decides right then, in that scroll, in those two seconds. The caption is not the hook. The video is.
Hooks are the first of the four mechanics HORN teaches — see the method. If you'd like yours looked at, send one video for a free teardown.