Free content hook generator for a stronger first draft.

Take Stonebridge Cycle Repair, a one person workshop in a railway arch making short vertical videos for people who commute by bike and service nothing until it fails. The video is forty seconds on brake pads. The audience is riders who have heard a squeal for a fortnight and done nothing about it, and the point is that they book a service before the pad wears into the rim.

The hook, in all three of its layers. First frame: a worn pad held up beside a new one, filling most of the screen, the groove worn completely flat, no face and no shop sign in shot. On screen text, four words, large: "This one is metal now." First spoken line, starting before the frame settles: "If your brakes have been squeaking for a couple of weeks, this is what is left of the pad, and the next thing it eats is your wheel rim." No greeting, no name, and no mention of the workshop until the final five seconds.

That holds up because the three layers work independently. Someone watching with the sound off gets the comparison and the four words, which is enough to understand the stakes. Someone who unmutes a beat later gets a sentence naming their exact situation, a fortnight of squeaking, in words they would use themselves. The object is on screen in the first frame rather than being introduced, so there is nothing to sit through before the substance arrives. The consequence is specific and expensive, since a wheel rim costs real money, which motivates far better than a vague suggestion of danger. And the payoff lands at around eight seconds, when the video shows how to check a pad in ten seconds without tools, so the promise is kept early rather than dangled.

The version the same shop filmed first: a wide shot of the arch, the mechanic walking towards the camera wiping his hands, saying "Hey guys, welcome back to the channel. Today I want to talk to you about something a lot of people ignore, which is brake maintenance. So stick around, because at the end I am going to share my top three tips."

Five seconds pass before any information appears, and short form retention falls off most steeply in exactly those seconds. The first frame is a room, which tells a scrolling viewer nothing and looks like every other workshop. Welcome back to the channel excludes everyone who has not been there before, which is nearly everyone the platform is testing it on. Something a lot of people ignore describes any topic at all. And stick around until the end asks for patience from someone who has not yet been given a reason to have any, which is the opposite of the trade a viewer is willing to make in the first two seconds.

If the video underneath has nothing in it, do not sharpen the opening. A hook is a promise about the next thirty seconds, and one that oversells produces a steeper drop at five seconds than a dull opening would have, plus a viewer less inclined to stop on your next post. When the honest summary of the piece is that it restates something obvious, the fix is a better video and not a better first line.

If people find the content by searching rather than scrolling, the calculation changes. Someone who typed how to fix a dripping tap has already decided to watch and wants orientation, so a tension based opening that withholds the answer irritates them. Search led and tutorial content is better served by confirming immediately that they are in the right place, naming the tool or the part, and saying how long the job takes. The same is true of longer form video where a viewer has committed several minutes: the first line still matters, but it works as orientation rather than as a snare.

Some subjects should not be hooked at all. Anything involving bereavement, illness, redundancy, a safety incident or a named individual's misfortune becomes distasteful the moment it is packaged for retention, and an audience notices instantly. Announcements with real consequences, such as a closure, a recall or a change affecting existing customers, should open by saying the thing plainly. So should anything a regulator would read, since a hook implying an outcome you cannot support is a claim regardless of how briefly it appears on screen.

And be wary of hooks that borrow attention rather than earn it. Fake urgency, a claim the video quietly abandons, engagement bait phrased as a question you never answer, or a shocking first frame unrelated to the content will all get watched once. What they cost is the follower who now treats the account as unreliable, and on platforms that weigh repeat viewing and follows earned from a video, that is a slower and more expensive kind of damage than a low view count.

A hook is not one sentence, it is three things arriving together, and a video fails when only one of them has been designed. The first frame carries the visual hook and is the only element some viewers ever see, because a proportion of the audience decides during the glimpse before playback really commits. It should show an object, an action already underway, or a state of affairs, rather than a person about to speak. On screen text is the second layer and carries sound off viewing, so it needs to be short enough to read in under a second, large enough to survive a small screen, and placed clear of the interface elements that overlay the bottom and right of a vertical video.

The spoken line is the third, and it should begin before the visual has fully resolved rather than after a beat of silence. Write the three separately and check each alone. Mute the video and ask whether the first two seconds still make sense. Then close your eyes, listen to the audio only, and ask the same question. If either fails, the hook is leaning on the other layer and half the audience is receiving a broken version of it.

The retention curve for short form video is close to useless as a single number and quite informative as a shape. Almost every video drops steeply in the first few seconds, so a fall there is normal; what matters is how steep it is relative to your own other videos, since absolute figures vary hugely by platform, by format and by how the video was distributed. Compare like with like, meaning your own posts of similar length and subject, rather than against a published benchmark drawn from accounts nothing like yours.

A drop concentrated in the first two seconds is a hook problem, usually a first frame that gave nobody a reason to stop. A drop at around five to eight seconds is a different fault: the hook worked and the payoff did not arrive, which is a structure problem in the body rather than in the opening. A flat curve with low reach means the hook is fine and the video was barely shown, which is a distribution question instead. Read watch time and completion together, because a very short video can complete at a high rate while holding nobody's attention in any meaningful sense.

They are often written by the same person in the same sitting, and they are not interchangeable. A headline is read at the reader's own pace, can be reread, and is judged beside competing headlines. A hook is heard once, at the platform's pace, with no chance to reconsider, and is judged against whatever the thumb was already moving towards. That single difference is why a headline can afford a subordinate clause and a hook cannot, and why a hook that reads well on paper often lands badly in the edit.

A thumbnail sits between the two and matters mainly where a viewer chooses from a grid rather than being served a feed. Its job is recognition at a glance, and it competes on contrast and legibility as much as on content. Where all three exist for one piece they should agree on the subject and differ in what they lead with, so the thumbnail shows the object, the title states the outcome, and the hook names the situation. Repeating the same sentence across all three throws away two of the three chances you get to interest someone.

Most weak hooks are repairable without reshooting, because the strong moment usually already exists further into the footage. Watch the raw file and find the first point where something is shown or stated rather than introduced, then cut everything before it. That single edit fixes a large share of openings, and the material you removed is nearly always greeting, setup and throat clearing that nobody asked for. If the best moment sits at twenty seconds, lift a two second fragment of it to the front as a cold open and let the video arrive at it again in sequence.

When a video underperforms and you believe the content is sound, recutting the first three seconds and posting again later is a reasonable experiment, particularly on platforms where an older post receives very little continued distribution anyway. Change one variable, note what you changed, and give it enough posts to mean something before drawing a conclusion. Keep a running list of the openings that worked on your own account, written as shapes rather than as sentences, since the specific words rarely transfer between subjects but the structure usually does.