On a short-video feed, people decide whether to keep watching almost immediately. They aren't judging your whole video. They're judging the opening line, the first frame, and the text on screen. If those don't give a reason to stay, the rest of the video never gets seen.
A hook isn't a gimmick. It's a promise: "keep watching and you'll get this." This post covers how to write that promise, six patterns that work for most topics, and how to test hooks instead of trusting your gut.
What a hook actually has to do
A good opener does three jobs in about three seconds:
- Say who it's for. "If you're learning to code…" filters in the right viewer and tells them this video is about them.
- Open a gap. A question, a tension, or a claim the viewer wants resolved.
- Fit what follows. If the hook promises "the one mistake" and the video is a general tips list, people feel tricked and leave anyway.
That third point is where most "clickbait" advice goes wrong. A hook that the video doesn't pay off loses viewers a few seconds later instead of right away.
Two hooks, not one: spoken and on-screen
Many people watch with the sound off, and the first frame often shows up as a still image in the feed. So treat every hook as two pieces that work together:
- Spoken line: what you say in the first breath. Short, conversational, no warm-up ("Hey guys, so today…" wastes your best seconds).
- On-screen text: 3–6 words that make sense without audio. It doesn't have to repeat the spoken line. It often works better as a label or a twist on it.
Example for a video about budgeting apps:
- 🗣 Spoken: "I tried five budgeting apps so you don't have to."
- 🖥 On-screen: "Only 1 was worth keeping"
The spoken line sets up the effort, and the text gives the reason to stay.
Six hook patterns (with examples)
Each pattern opens the gap in a different way. Each example below shows the spoken line, then the on-screen text, for a made-up topic.
1. The mistake
Tell viewers they're doing something wrong, then promise the fix.
- 🗣 "You're not bad at sleep. You're doing the last hour wrong."
- 🖥 "The last-hour mistake"
It works because people want to know if they're the one making the mistake.
2. The contrarian claim
Go against common advice, but only if you can back it up in the video.
- 🗣 "Stop writing a to-do list every morning."
- 🖥 "Do this instead"
3. The specific question
A vague question ("Want to be more productive?") gets skipped. A specific one gets watched.
- 🗣 "Why do most people give up on Duolingo around day 40?"
- 🖥 "The day-40 wall"
Only use a number like this if you have a source for it, or put it to the viewer as a question ("Ever notice you quit around week six?").
4. The result first
Show the end state in the first frame, then go back and explain how.
- 🗣 "This desk setup cost less than a single monitor arm."
- 🖥 "Full breakdown →"
5. The callout
Name the viewer directly. This narrows your audience on purpose.
- 🗣 "If you freelance and hate invoicing, this one's for you."
- 🖥 "Freelancers only"
6. The countdown or list
A number tells people how much they're signing up for.
- 🗣 "Three settings to change on a new iPhone, and the third one matters most."
- 🖥 "#3 is the one"
Telling them the last item matters most is a reason to stay to the end.
A quick checklist before you film
Run each candidate hook through these questions:
- Could I say it in one breath? If not, cut it.
- Does the on-screen text make sense on mute?
- Is there a specific noun or number? "Productivity tips" is weak, "the 2-minute rule for email" is better.
- Does the video deliver it in the first 10–15 seconds? Don't hold the payoff until the end if the hook promised it.
- Is it true? Made-up stats and "doctors hate this" lines cost trust over time, and on some platforms they can get your reach limited.
Test hooks, don't guess
The most useful habit is to write several hooks in different patterns for the same video and treat the opener as the thing you're testing. Some ways to do that:
- Post the same core video with two different openers on different days or accounts and compare how long people watch.
- Use the platform's own A/B or test features if you have them.
- Keep a simple sheet: topic, hook pattern, hook text, how long people watched. After 10–20 posts you'll see which patterns work for your audience.
Five versions of the same pattern isn't a test. A mistake hook against a question hook against a result-first hook tells you something useful.
Then script around the winner
Once you've picked a hook, the script's job is to keep the promise quickly. A simple shape for a 30-second video:
| Time | Beat |
|---|---|
| 0–3s | Hook (spoken + on-screen) |
| 3–10s | Name the problem or tension |
| 10–22s | The payoff: the fix, steps, or reveal |
| 22–30s | One call to action (follow, save, comment, link) |
Keep to one idea per video. If you have three ideas, that's three videos with three hooks.
Automating the first draft with Hook Lab
Writing five or more hooks in different patterns, picking one, and scripting around it takes time every time you film. My team built an AI agent for this step called Hook Lab. It runs in a chat on Capafy, an AI agent marketplace.
What you send: a topic, product, trend or niche, plus the platform (TikTok, Reels or Shorts) and the video length if you know it.
What you get back:
- a positioning angle, meaning the point of view your account should own on that topic,
- 5+ first-3-second hooks, each with a spoken line, on-screen text, and the hook type it uses, so you can A/B them,
- a recommended pick and a beat-by-beat script with timecodes for your length,
- a caption with a call to action, hashtags, and one filming tip.
For example, the listing shows a request for "Hooks for a 30s TikTok about why most people quit the gym in January — energetic, for beginners." Two of the hooks it returns are a contrarian one ("Stop going to the gym every day." / "Day 1 is where you quit") and a mistake one ("You're not lazy. You're doing week one wrong." / "Week 1 mistake"), followed by a 0–3s / 3–10s / 10–22s / 22–30s script.
What it doesn't do matters too. It works only from the text you send. It doesn't watch your videos, log in to your accounts, post anything, or read your analytics. You can ask for changes in the same chat ("make #2 funnier", "shorter script").
Pricing on the listing is $3.99/day, $9.99/week or $19.99/month. The listing suggests the day plan for a one-off batch and week or month if you film regularly. Treat what it gives you the way you'd treat any first draft: check that the hooks are true for your video and match how you talk.
Wrap-up
- The first three seconds are a promise. Make it specific and keep it.
- Write a spoken line and on-screen text that works on mute.
- Use different hook patterns so your tests tell you something.
- Script for a fast payoff, one idea per video.
Whether you write them by hand or start from a generated draft, the habit that helps most is testing different hook patterns against each other.
Disclosure: my team built Hook Lab and sells it on Capafy.
Top comments (0)