SHORT-FORM VIDEO HOOK PROMPTS

4 primary sources

7 Short-Form Video Hook Prompts You Can Copy and Use

Choose a prompt by what your opening needs to do: break a pattern, raise a question, name a problem, show proof, or begin a story. Fill in your audience, actual payoff, and available shots. Each prompt returns spoken words, on-screen text, a first frame, and where the video delivers the promise. Two follow-up prompts help you rewrite an existing opening and plan a fair comparison.

Written and researched by Kabo Editorial TeamReviewed by Emma HsiaPublished October 23, 202619 min read

01 · START WITH THE JOB

Choose the job your opening needs to do

You already know what the video is about. The awkward part is finding an opening that sounds like you and gives someone a reason to stay. Asking for ten “viral hooks” often leaves you choosing between ten versions of the same big promise. A smaller request is easier to judge: what can this video show, and which opening makes that worth watching?

What you have	Try this prompt	Check before filming
An unexpected contrast	Pattern break	The contrast is true and visible.
A question with a useful answer	Curiosity gap	The video actually answers it.
A recognizable problem and a fix	Problem-solution	The benefit fits the demonstration.
An inspectable result	Proof-first	The conditions behind the result remain clear.
A real event with a turning point	Story	The experience belongs to you.

These are writing choices, not platform ranking categories. TikTok's Creative Codes recommends a hook, body, and close for ads, with devices such as surprise or suspense. That is useful inspiration for an opening, but it does not establish a winning formula for every organic post.

TikTok for Business: Creative Codes

If the first sound does the work, pair the chosen prompt with the guide to using sound as a video hook. A click, a pause, or a spoken question still needs to belong to the scene you are making.

02 · GIVE THE MODEL SOMETHING REAL

Fill in the facts before you ask for hooks

“A video about desk setups” gives the model room to invent the interesting part. “I can show glare disappearing when I move this lamp” gives it an opening you can film. The prompts below ask for that difference: a viewer, an actual payoff, supporting facts, and shots you can use.

Add a couple of lines you wrote yourself. They give the model something more useful than “make it authentic.” A clear output format also makes the answer easier to check; Anthropic's prompting guidance recommends explicit context, format, and constraints. OpenAI's guide also recommends clear instructions, relevant context, and examples of the desired output.

Anthropic: prompting best practices OpenAI: prompt engineering

Each copy button includes the whole prompt, so you can use one without copying a separate setup. Replace the bracketed fields with your notes. Write NONE when you have no supporting result or number. The model should ask for missing essentials instead of turning a blank into a testimonial.

For the opening budget, use the delivery time you can actually say aloud and the amount of text you can comfortably read on a phone. If you need help setting that budget, see hook length and word-count examples. The purpose here is to generate usable options within your budget, then rehearse them.

03 · COPY THE TASK YOU NEED

Copy five hook-generation prompts

Paste one prompt into a text assistant you already use. All five ask for the same opening package, which makes the options easier to compare. The first frame is a filming plan; a text answer does not create the footage.

1. Interrupt an expected opening

You have a concrete contrast to show, but your current opening explains the topic before showing anything.

PROMPT 1 · PATTERN BREAKInterrupt an expected opening

Help me write an opening I can actually film.

TASK
Create a pattern break using an unexpected but truthful contrast in the supplied material. Pair each opening with an available visual action. Explain the expectation it interrupts. Avoid fake controversy, insults, and unsupported 'everyone is wrong' claims.

OUTPUT
Return three meaningfully different candidates labeled A, B, and C.
For each, give:
1. Spoken opening, in my voice.
2. On-screen text that makes sense without the audio.
3. First frame and the action visible in it.
4. The exact later beat that delivers the promise.
5. Which supplied fact or shot supports the wording.
Finish with your clearest candidate and why it fits this video. Rank by clarity and feasibility, not predicted views.

RULES
- Use only the supplied facts, experiences, and available shots.
- Do not invent numbers, results, testimonials, expert authority, or a personal story.
- If an essential fact or shot is missing, ask for it instead of filling it in.
- Describe shots as plans; do not claim to have filmed or edited them.
- Keep the opening within my stated budget; flag timing for a read-aloud check.
- Do not promise a payoff missing from the video or guarantee performance.

The interruption should come from your material. A louder adjective will not create a new shot.

2. Raise a question the video answers

The payoff is useful, but revealing it immediately would remove the reason to watch the demonstration.

PROMPT 2 · CURIOSITY GAPRaise a question the video answers

Help me write an opening I can actually film.

TASK
Create a curiosity gap around one specific question the supplied video answers. Give enough context for the viewer to understand the question. Name the answer and its payoff beat separately from the opening. Do not withhold a qualification that changes the claim's meaning or delay the answer just to stretch the video.

A specific unanswered question gives you something to resolve. 'You will not believe this' gives you another line to rewrite.

3. Name the viewer's problem

You can demonstrate a useful fix for a problem your intended viewer already recognizes.

PROMPT 3 · PROBLEM-SOLUTIONName the viewer's problem

Help me write an opening I can actually film.

TASK
Open with a recognizable problem in the viewer's stated situation, then point toward the specific solution demonstrated here. Keep the problem and benefit proportional to the supplied facts. Show the symptom in the first frame when an available shot supports it. Do not invent an audience survey, universal cause, or guaranteed result.

Keep the fix as small as the video. Showing a better lamp position does not support a promise to transform someone's whole setup.

4. Show the result before the explanation

Your actual footage contains a result or side-by-side comparison the viewer can inspect.

PROMPT 4 · PROOF-FIRSTShow the result before the explanation

Help me write an opening I can actually film.

TASK
Lead with a result or comparison visible in the supplied shots. Keep any qualification needed to interpret it beside the claim. Point to the supporting fact and shot for every result statement. If the material contains no inspectable result, say proof-first is not supported and suggest another opening task. Never turn one demonstration into a universal performance claim.

Proof-first is a filming choice, not permission to remove the conditions behind a result.

5. Begin at a real turning point

You have a real sequence of events with a decision or discovery worth following.

PROMPT 5 · STORYBegin at a real turning point

Help me write an opening I can actually film.

TASK
Begin at a turning point from the supplied real events. Use first person only for experiences explicitly present in the input. Make the immediate situation understandable, and name the later beat that resolves it. If no personal events were supplied, ask for them or propose a clearly labeled hypothetical scenario. Do not present a hypothetical as my life story.

A story needs an event. If you only have a tip, the problem-solution prompt may fit better.

04 · FROM ANSWER TO A SHOT

See a filled example before you film

Constructed example. Imagine you are filming a desk-lamp positioning tip. The footage described below is hypothetical.

Your viewer works at a small desk and sees a bright reflection on a laptop screen. You can film the same screen with the lamp in two positions, keeping the screen brightness and camera exposure fixed. The payoff is modest: show how the reflection changes in this setup. You have no eye-strain measurements, energy savings, or product test results.

Here are two possible packages using that input. The first names the problem; the second lets the comparison lead. Either could fit the same body, where you show both positions and point out what changed.

Opening	Spoken line and text	First frame and payoff
Problem-solution	“That bright patch? Try moving the lamp.” Text: “Glare in this setup”	Open on the reflection. Move the lamp in view, then show both positions.
Proof-first	“Same screen. Different lamp position.” Text: “Watch the reflection”	Open on the two-position comparison. Explain the fixed settings before demonstrating the move.

A line such as “This trick cuts eye strain in half” would need evidence you do not have. Reject it even if it sounds punchier. The useful option makes the visible change clear without upgrading a reflection demo into a health claim.

Read the chosen line aloud while looking at the proposed first frame. If the words describe one thing and the shot shows another, fix the package before filming. You should be able to point to the moment later in the video that fulfills it.

05 · KEEP THE IDEA, FIX THE OPENING

Rewrite an opening that sounds generic

Sometimes the idea is fine and the line is doing too much setup. “Today I want to share a useful desk tip” leaves the viewer waiting to find out which tip and why it matters. A rewrite can bring the visible problem forward without changing the rest of the video.

PROMPT 6 · REWRITEKeep your meaning while making the opening specific

Help me rewrite an opening without changing what the video promises.

INPUTS
- Platform: [TikTok, Instagram Reels, or YouTube Shorts]
- Viewer and their situation: [who is watching and what they need]
- Video topic: [one specific idea]
- Actual payoff: [what this video will show or explain]
- Supporting facts: [verified facts, numbers, and qualifications; write NONE if absent]
- Available shots: [what I can show, with file names if already filmed]
- My voice: [two lines I wrote, plus phrases I would never say]
- Opening budget: [my spoken delivery time and readable text limit]
- Remaining video: [outline or script, including where the payoff appears]
- Current spoken opening: [paste verbatim]
- Current on-screen text and first frame: [describe]
- What feels wrong to me: [vague, stiff, confusing, or another specific concern]

TASK
Identify the unclear words, unnecessary setup, or mismatch with the first frame. Do not diagnose audience behavior from the wording alone.
Write three alternatives:
A. A minimal rewrite that preserves my phrasing.
B. A more specific version supported by my facts.
C. A different opening task suited to my available shots.

Choose the smallest rewrite that solves the problem. Read it against your footage and restore any qualification the model dropped. If the answer invents a result, correct the input before asking for another round.

That generator writes a script from a description. To use these exact prompts and compare their structured answers, paste them into your text assistant.

06 · MAKE THE DIFFERENCE READABLE

Plan a comparison you can learn from

Two attractive openings can differ in everything at once. In the lamp example, switching the words, overlay, and opening shot would leave you unsure which change mattered. For a first comparison, you could keep the reflection shot and text the same, then compare a question with a direct instruction.

PROMPT 7 · COMPARISON PLANChange one part of the opening

Help me plan a manual comparison of two openings. Do not post, collect data, or claim a winner.

INPUTS
- Platform: [TikTok, Instagram Reels, or YouTube Shorts]
- Viewer and their situation: [who is watching and what they need]
- Video topic: [one specific idea]
- Actual payoff: [what this video will show or explain]
- Supporting facts: [verified facts, numbers, and qualifications; write NONE if absent]
- Available shots: [what I can show, with file names if already filmed]
- My voice: [two lines I wrote, plus phrases I would never say]
- Opening budget: [my spoken delivery time and readable text limit]
- Remaining video: [outline or script, including where the payoff appears]
- Current opening package: [spoken line, on-screen text, first frame]
- Proposed opening package: [spoken line, on-screen text, first frame]
- Metric available in my platform analytics: [exact label and definition, or UNKNOWN]
- My usual observation window: [when I will record the results]
- Constraints: [what I can realistically keep the same]

TASK
Choose one difference to compare: spoken framing, on-screen text, OR first-frame action. If the packages differ in several ways, revise them so only the chosen variable changes. Preserve the same payoff and qualifications.

OUTPUT
Return a plan with:
- One hypothesis stated as a question.
- Version A and version B, each with spoken line, text, first frame, and payoff beat.
- What stays the same and what changes.
- A metric I can actually read, with a separate check that the video delivers the promise.
- The observation window I supplied and a blank results log.
- Audience, timing, distribution, or sample differences that could explain a result.
- When to leave the result inconclusive and what to try next.

RULES
- Use the supplied facts and shots only; ask for missing essentials.
- Do not invent analytics, minimum winning thresholds, or a guaranteed reach increase.
- If the metric is UNKNOWN, ask me to check my analytics before selecting one.
- Separate a controlled platform experiment from separate organic posts.
- Treat an organic comparison as a clue for the next test, not proof that the hook caused the outcome.

This gives you a plan and a blank log. You still choose how to run the comparison and enter your own results. If your platform offers no controlled test for that format, separate organic posts will reach different people at different times.

For YouTube Shorts, Stayed to watch describes how often viewers watched versus swiped away. Average view duration describes time watched among those who stayed. Those answer different questions: did people stay for the opening, and did the video keep delivering afterward? Use the exact metrics available in your own dashboard rather than applying one platform's labels to another.

YouTube Help: understand your content performance

Record both results after the same observation window and note changes in audience, topic, or distribution. One higher view count is a reason to investigate, not a verdict on the words. An inconclusive comparison can still show which question to test next.

Once a cut has a public link, the AI Video Analyzer can give timestamped feedback on the opening and payoff. It reads the public video, not your private retention report. Use that feedback to spot a confusing shot or an unfulfilled promise, alongside the results you can see in your platform analytics.

SOURCES

Sources

Anthropic's and OpenAI's documentation informs the prompt structure. Results will vary with the model you use and the details you provide. Try them with your own material and check the wording and promised payoff. TikTok's Creative Codes is guidance for advertising creative. YouTube's help page defines the Shorts metrics discussed above. The task categories, review questions, and comparison plan are Kabo's editorial recommendations; the lamp example illustrates a writing decision, not a measured performance result.

Source guidance checked 2026-10-12.