The first three seconds are not a figure of speech. A short-form feed is full screen, autoplaying and one flick deep, so every video gets exactly one audition: the moment between appearing and being swiped past. Everything you did after that moment, the edit, the grade, the argument, only exists for the people who stayed. The hook is the part of the video that earns them.
That makes the hook the highest-leverage thirty frames you will make all month, and also the most commonly skipped. Most videos do not fail because their hook was weak. They fail because there was no hook at all, just an intro.
What a hook actually is
A hook is a specific promise with a deadline. It tells the viewer what they will get if they stay, and it makes that offer before the swipe reflex fires. Vague curiosity is not enough; “wait for it” is a promise with no contents, and feeds are full of it.
It also arrives on three channels at once: the first frame (what you see), the first line (what you hear), and the text on screen (what you read with the sound off, which is how a large part of any feed is watched). The three should make the same promise. A strong spoken line over an establishing shot of your office is a hook cancelled by its own footage.
A hook is a promise with a deadline. You have three seconds to make it and the rest of the video to keep it.
None of this is a trick, and it does not require shouting. It requires deciding, before you make the video, what the one thing in it is, and then leading with that instead of walking up to it.
Six shapes that keep working
Hooks are a small genre and the reliable shapes are known. If you are staring at a blank script, start from one of these:
- The question the viewer already has. Say it in their words, then answer it. The video feels like a reply rather than an ad.
- Starting mid-action. The product already in use, the result already on screen. Context can wait; motion cannot.
- The objection, head on. “This looks like it would break in a week” is a stronger opener than any compliment, because doubt is what the viewer was thinking anyway.
- The before state. Show the problem in its natural mess. A viewer who recognises the before will stay for the after.
- The specific number. Your real price, your real time saved, your real count of attempts. Specifics read as evidence; round claims read as copy.
- The claim you can defend. One sharp sentence you are prepared to spend the whole video backing up. It filters for exactly the viewers you want.
Notice what is not on the list: your logo, your founding story, and the word “introducing”.
Test hooks instead of debating them
The uncomfortable part is that you cannot reason your way to the right hook. Taste arguments about openers are how teams spend a meeting deciding between two versions that both lose. Retention data settles it in a day.
The mechanics are simple. Take one video and make several versions that differ only in the first three seconds, post them, and compare how many viewers survive the open. That figure is the hook rate, and the hook rate calculator turns any platform’s numbers into it so versions can be compared like for like. Everything else in the video stays constant, so whatever difference you see is the hook.
This used to be an expensive experiment, because five openers meant five shoots. Generated video removed that constraint: one prompt can produce a deck of variants in minutes, which makes hook testing a routine habit rather than a quarterly project. The volume argument from how often to post applies doubly here; five hook tests a month is the difference between having an opinion about your audience and having evidence.
The habits that flatten a hook
A few patterns reliably spend the three seconds on nothing. Logo stings and channel intros, because they promise a brand rather than a payoff. Greetings, because “hey everyone, welcome back” is three seconds of content-free audio. Saving the reveal for the end, which works in cinema and dies in a feed. And asking for a follow up front, which is asking for payment before the product.
The fix for all of them is the same edit: find the most interesting second of the video and move it to the front. If you are picking formats from a library, the Templates gallery shows how each shape performed, which is a shorter path than inventing an opener from scratch. But whatever you start from, the discipline holds: make the promise first, keep it fast, and let the data pick the winner.
