Twelve opening lines, each labelled with the pattern it came from and measured against your format's limit. Type a topic and an audience, then pick a format and one emotion lever.
.png)
Twelve named patterns; your topic and audience drop into forty eight written templates right in your browser, and nothing you type is sent or kept. Any number inside a hook is a placeholder, not a measurement, so replace it with yours before you post.
How long do you really have to hook a viewer?
An Instagram hook is the opening line of a Reel, carousel or caption, the words that tell someone what they are about to get before they decide whether to stay. Every guide gives you three seconds to land it. Meta has never published a study behind that number.
We checked the eight most popular hook generators and hook lists in August 2026. All eight repeat a first seconds claim. None links a primary source for it. None labels the pattern on the hook it hands you. One gives a length rule. This tool builds every line from twelve named patterns, prints the pattern, and measures the line against the format you picked.
How this generator works
How the twelve hooks are built
The hook matrix
Example:
Topic meal prep, audience busy parents, format Reel, lever Curiosity. The first line comes back as "Don't touch meal prep until you've seen this." labelled Negative command and Curiosity, 45 characters and 8 words, marked as fitting one spoken breath. Ten of the twelve fit that rule. The Objection and Ranking lines, both 13 words, are marked too long to say in one breath.
No study produced these twelve patterns. What each rule rests on:
Your topic is the input everything hangs on: change it and all twelve lines change with it.
Why a labelled hook beats a generated one
Where the three second rule comes from
Four sources get welded into one rule of thumb. Only the last is about hooks.
| Source | Date | What it actually says | Is it about hooks? |
|---|---|---|---|
| Meta Business Help Center, 3-second video plays | August 2026 | A play counts when a video "played for at least three seconds, or for nearly its total length if it's shorter than three seconds" | No. It is a counting rule for advertisers |
| Facebook, The Value of Video for Brands, Nielsen, 173 BrandEffect studies | 17 March 2015 | "people who watched under three seconds of the video ad created up to 47 percent of the total campaign value" | No. It argues short views already count |
| Facebook IQ, Capturing Attention in Feed | 20 April 2016 | "0.25 seconds of exposure for people to recall mobile feed content" and "1.7 seconds with a piece of content on mobile" | No. The phrase three seconds never appears in it |
| Adam Mosseri, Head of Instagram | 4 October 2024 | "The first three seconds of a reel are super important" and the best hooks "use text and set expectations for what you're about to see" | Yes, and it gives no number |
Both Facebook pages no longer load; the two studies everyone cites survive only in the Internet Archive captures linked above.
Handling DMs by hand vs automating with Inrō
A hook that oversells earns one scroll now and teaches your audience to skip the next thing you publish.
The templates carry invented figures, and a number you cannot back turns your comment section into a cross examination.
Unlabelled hooks tell you that a line worked. Labelled ones tell you which pattern worked, which is the reusable part.
Run the curiosity lever on Reels and keep the pattern label, so your swipe file records shapes rather than one off lines.
Stakes works when your audience already knows the problem, so name what it costs rather than teasing the answer.
Specificity suits product posts, but swap the placeholder figures for your real numbers before anything goes near a caption.
Contrarian gives a client account a voice, so run it across three posts before you decide it fits the brand.
Nobody at Meta has published a study showing you lose a viewer at second four. What the Head of Instagram did say, on 4 October 2024, is that hooks which set an expectation work, and that you should test different types. Twelve named patterns is where testing starts.
Turn the hook into a conversation
A hook that lands fills your comments. Set a keyword trigger and every one of those comments gets a DM in seconds. The free plan covers 100 active contacts a month, 3 automations, 3 campaigns and 50 AI credits.
Frequently asked questions
A hook does one job. It tells someone what they are about to get, before they decide whether to keep going. Adam Mosseri, the Head of Instagram, put it in a reel on 4 October 2024: the best hooks he has seen "use text and set expectations for what you're about to see". That is the whole brief. A hook that sets an expectation the post then meets earns the watch. One that promises something the post never delivers earns a scroll and teaches your audience to skip the next one. The twelve patterns here are twelve ways to set that expectation.
Three separate things, and none is a study of hooks. Three seconds is Meta's ad counting threshold: its Business Help Center defines a 3-second video play as a video that "played for at least three seconds, or for nearly its total length". The 47 percent figure everyone quotes comes from Facebook's "The Value of Video for Brands", 17 March 2015, where Nielsen analysed 173 BrandEffect studies and found that people who watched under three seconds still created up to 47 percent of campaign value. That is an argument that short views count, not a deadline. Mosseri then said on 4 October 2024 that the first three seconds are "super important", with no number attached to it.
There is no published limit, so this tool judges every hook against the surface you picked and names the rule it is applying. A Reel hook is scored on words, twelve or fewer, because it gets spoken in roughly one breath. A carousel hook is scored on characters, sixty or fewer, because slide 1 is often first seen at thumbnail size. A caption first line is scored against 125 characters, where guides say the feed folds the rest away. The first two are our own rules. Only the 125 figure comes from anywhere else.
No. Every hook comes from a written template, and the same four inputs always return the same twelve hooks in the same order. Twelve named patterns and four levers make forty eight templates, and your topic and audience drop into them unchanged. Nothing is sent to a model, nothing leaves your browser and nothing is stored. That is a choice rather than a shortcut. An AI hook writer will hand you a line with a number in it that you never earned, and nothing tells you the number was invented. A template that prints its own pattern name is easier to check.
A pattern is the shape of the sentence. A lever is the reason anyone would care about it. Question, Open loop, Numbered list, Negative command, Comparison, Confession, Direct address, Conditional, Overheard, Time frame, Ranking and Objection are shapes. They describe grammar, not feeling. Curiosity, Stakes, Specificity and Contrarian are levers. They describe what the sentence withholds, what it threatens, what it quantifies or what it contradicts. Keeping the two apart is the point of the tool. Most hook lists mix them, so a curiosity gap and a question end up as siblings when one is a motive and the other is a grammar. Change the lever here and all twelve shapes stay put while every sentence gets rewritten.
Because they are. The Specificity lever produces lines such as "It takes eleven minutes" and "The best advice I ever got was four words long". Those numbers come from the template, not from your work, and posting one unchanged would be a claim you cannot back. The widget says so in its own note instead of hiding it. This is the failure mode of AI hook writers that is easiest to miss. They emit confident specifics because specifics perform, and the interface never flags which specific was invented. Replace every number in a hook with one you can point at. A line carrying your real figure beats a borrowed one, and it is the only version you can defend in the comments.
Both, and they do not have to be the same words. Mosseri's reel of 4 October 2024 singled out hooks that "use text and set expectations", which is a vote for putting it on screen. In practice the text hook has to survive being read with the sound off, and the spoken hook has to survive being heard while someone glances away. That is why this tool scores a Reel hook on words rather than characters. Twelve words is roughly one spoken breath, a constraint the on-screen version does not have. If only one of the two can carry it, make it the text.
It is free, there is no account, and there is no email step in front of the results. The whole tool runs inside this page. Your topic, your audience, your chosen format and your lever never leave the browser, nothing is written to a server and nothing is kept between visits. Reload the page and it returns to the example it shipped with. The copy button uses your own browser clipboard. Once you have the opening line, the Instagram Caption Generator on this site builds the three parts that come after it and marks the point where the feed cuts a caption off, so you can see how much of the rest survives the fold.
More than one, which is the only instruction the platform itself has given. Mosseri's reel of 4 October 2024 closed with "try experimenting with different types of hooks to see what works best for you", and that is as specific as Instagram has ever been in public about hooks. Nobody publishes a number, so this page will not invent one. What the tool gives you is twelve at a time, each labelled with the pattern it came from, so when one of them lands you know which shape landed and can write more of that shape yourself. Testing unlabelled lines tells you that a line worked. Testing labelled lines tells you why it worked.
Attract more leads, reply in seconds, and turn engagement into sales, automatically.
