Free AI Video ToolsFree AI Video Tools
Free AI video hook prompt library

Go Viral with TikTok & Instagram Hooks.

A library of 3-second hook prompts, the opening beat that decides whether your video gets watched. Pick a format, take its prompt, and render the hook your Reel or TikTok is missing.

  • Pick a hook format, nothing to upload
  • Make the line your own
  • Render it vertical in seconds
Browse the hooks

Free. No signup. Works with any AI video model.

Impossible shot

Giant head held still by a hand while the tiny body spins beneath

Pattern break

Ceiling bursts open and a body crashes down through the debris

Impossible shot

Presses on the glass from inside the TV, then climbs out into the room

The hook library

Real hooks,
three seconds each

Every card is one opening beat, tagged with the job it does in the first second. Where the prompt behind a clip is written, hover the card (or tap Prompt on mobile) to read it and take it into Studio AI.

Impossible shot
A photorealistic, high-angle shot looking down at a human subject. The subject has an extreme caricature distortion applied: a disproportionately massive, oversized head attached to a tiny, miniature body. A large hand enters from the top edge of the frame, firmly holding the top of the subject's giant head. While the massive head is held completely still and facing forward, the tiny miniature body rapidly spins around in continuous 360-degree circles beneath it. Soft, natural lighting. Crucial instruction: Absolutely no text, typography, watermarks, subtitles, or UI elements anywhere on the screen.

Hand holding giant head

Giant head held still by a hand while the tiny body spins beneath

Pattern break
A dynamic, low-angle shot inside a room. The ceiling suddenly shatters and bursts open as a human subject violently falls through a jagged hole from above onto a couch. A explosion of thick pink dust, smoke, and heavy debris cascades down alongside them from the realistic isolation material in the ceiling. Real speed, no slow motion. The subject crashes heavily onto a piece of furniture below, landing in a dynamic crouching pose. The impact causes a sudden camera shake. Crucial instruction: Absolutely no text, typography, watermarks, subtitles, or UI elements anywhere on the screen.

Human falling through ceiling couch

Ceiling bursts open and a body crashes down through the debris

Impossible shot
A one shot, static shot of a room with a large flat-screen television in a realisitc living room. A human subject is initially inside the TV's display, wiping the inside of the screen. The subject pushes against the glass, and the screen surface suddenly ripples and stretches outward like liquid. The subject physically climbs completely out of the television screen and into the real room, standing up with a shocked expression. Crucial instruction: Absolutely no text, typography, watermarks, subtitles, or UI elements anywhere on the screen.

Person exiting television screen

Presses on the glass from inside the TV, then climbs out into the room

Pattern break
A dynamic, medium-full shot of a human subject standing indoors at eyes level. The subject looks up in anticipation just as the ceiling above them suddenly cracks and violently bursts open. A massive, heavy torrent of water crashes down like a waterfall directly onto the subject. The immense volume of water completely drenches them instantly, visibly soaking their clothes and hair. The subject reacts with a dramatic, wide-mouthed expression of intense shock, throwing their arms out slightly to the sides as the heavy deluge continues to pour over them and splash onto the floor. Crucial instruction: Absolutely no text, typography, watermarks, subtitles, or UI elements anywhere on the screen.

ceiling breaking

Ceiling gives way and dumps water over her, open-mouthed, drenched

Impossible shot
A surreal, medium-full shot of a human subject standing indoors. The subject's body is entirely headless. Their head is massively oversized, physically detached from the body, and floating high up in the air beside them exactly like a giant helium balloon. The headless body holds a thin string that is visibly tethered to the bottom of the floating, oversized head styled like a balloon. The giant head bobs and drifts gently in the air, kept from floating away solely by the string held by the body's hand. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

floating head balloon

Her head floats above her like a balloon, tethered on a string

Impossible shot
A surreal, static, full-body shot of a human subject. The subject is in a regular living room posed in a seated position with their legs crossed and feet touching the floor, but instead of sitting on a piece of furniture like a small stool, they are sitting directly on top of a massive, oversized, disembodied replica of their own head. The giant head rests flat on the floor, functioning as a chair and supporting the subject's weight. The giant head is alive and expressive, blinking its eyes and smiling slightly, while the actual subject sits calmly on top of it looking forward. the size of the sitting subject should be relative to the size of the furniture around it so the head is the same size of a chair and the human sitting size is relative to other furniture, hence, making it look real Crucial instruction: Absolutely no text, typography, watermarks, subtitles, or UI elements anywhere on the screen.

Person sitting on giant head

A miniature of herself sits cross-legged on top of her own head

Reaction
hyperrealistic phone recorded style close-up, frontal shot of a human subject sitting in a living room in a couch. Another person's hand swiftly enters from the side of the frame and slaps the subject across the cheek at real speed. The intense impact is captured with exaggerated realism, causing the subject's facial features to distort heavily, their skin to ripple, and their hair to fly wildly in the direction of the strike. The subject's eyes squeeze shut from the force. Slight camera shake upon contact. Crucial instruction: Absolutely no text, typography, watermarks, subtitles, or UI elements anywhere on the screen.

face slap

Deadpan to camera, then a hand swings in and the face folds

Reaction
An extreme close-up, front-facing shot of a human subject's face. The subject has one hand clamped firmly over their mouth in a dramatic gesture of intense, joyful shock or suppressed laughter. Their eyes are wide open and staring directly into the camera lens, with their eyebrows raised high in delightful surprise. The strong emotion of astonishment and excitement is heavily conveyed through their wide, expressive eyes and the subtle crinkling of the skin around them, while their mouth remains entirely hidden behind their hand. The camera is completely steady, maintaining a tight, intimate focus on the subject's highly expressive face. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

shocked face

Hand clamped over her mouth, eyes wide straight down the lens

Reaction
A static, medium-full shot of a human subject sitting in a black office chair at a computer desk. The subject is initially hunched forward, burying their face deep into their hands in a posture of extreme stress, frustration, and defeat. Suddenly, their demeanor completely flips. They abruptly pull their hands away from their face, throwing them up to grab their head and hair in disbelief. They look up with a sudden, ecstatic expression of extreme joy, relief, and shock, breaking into a massive, wide-open smile. Their body language instantly shifts from depressed and defeated to wildly triumphant and energetic. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

male reaction

Head buried in his hands at the desk, then flips to open-mouthed triumph

Corrective
A close-up, eye-level shot of a human subject actively speaking directly to the camera. The subject's lips are moving dynamically, clearly mouthing words and articulating a sentence with an engaging, instructional expression. the subject says " Dont do X, do Y instead" The subject holds a pen in one hand, gesturing with it slightly to punctuate their speech. Mid-sentence, they bring their other hand into the frame, decisively pointing their index finger straight up to emphasize a key point. The subject's head and body make subtle, natural conversational movements that synchronize perfectly with the act of talking. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

dont do X, do Y

Holds the product up, points at it, corrects you

Corrective
A close-up, eye-level shot of a human subject actively speaking directly to the camera. The subject's lips are moving dynamically, clearly mouthing words and articulating a sentence with an engaging, instructional expression. the subject says " Dont do X, do Y instead" The subject holds a pen in one hand, gesturing with it slightly to punctuate their speech. Mid-sentence, they bring their other hand into the frame, decisively pointing their index finger straight up to emphasize a key point. The subject's head and body make subtle, natural conversational movements that synchronize perfectly with the act of talking. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

dont do X, do Y instead

Podcast mic, glasses, delivers the correction to camera

Corrective
A close-up, slightly low-angle shot of a human subject indoors, looking directly into the camera lens with a steady, neutral expression. The subject raises one hand, pointing their index finger straight up. They smoothly and deliberately sweep their pointed finger horizontally across the foreground from one side to the other, and then slowly back again, performing a continuous "no" or wagging gesture. The subject maintains unwavering eye contact throughout the fluid motion. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

X + Y = Z

Wags a raised finger slowly side to side, never breaking eye contact

Pattern break
A close-up, side profile shot of a human subject indoors. The subject is holding a massive, clear plastic gallon jug filled with white milk up to their face, with the opening positioned right at their lips as if taking a huge gulp directly from the container. Suddenly, the subject abruptly yanks the heavy milk jug down and away from their mouth. Simultaneously, they quickly turn their head forward, looking straight ahead with a frozen, wide-eyed expression of shock and surprise, as if they were just caught in the act. The sudden, guilty movement creates a brief, dynamic shift in their posture. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

drinking milk

Mid-chug from the gallon jug, then snaps to camera like he got caught

Silent action
A dynamic, static medium shot of a human subject standing at a kitchen counter, leaning over a wooden cutting board. The subject is holding a large chef's knife and furiously chopping a small item. The chopping motion is hyper-exaggerated, frantic, and unnaturally fast, creating a continuous blur of rapid kinetic movement. The subject's upper body, arms, and head bob violently and aggressively up and down in perfect rhythm with the rapid-fire knife strikes, causing their hair to shake and flop wildly. The camera angle is slightly elevated, looking down at the countertop to emphasize the intense, obsessive speed of the chopping action. Bright, even, natural indoor lighting. Crucial instruction: Absolutely no text, typography, letters, words, watermarks, subtitles, or UI elements anywhere on the screen.

cutting

Chopping at an absurd, frantic speed, whole body bobbing with the knife

Renders in the best AI video models

Veo 3.1 Kling 3.0 Runway Gen 4.5 SeeDance 2.0 Luma Ray 3.2 Pika 2.1

Most hooks carry a spoken line, so the prompts open with Veo 3.1 pre-selected for its native audio and lip-sync.

How it works

Fix your first
three seconds

Nothing to upload and nothing to film. Pick the opening beat you need, make the line yours in Studio AI, and put it in front of the video that isn't holding attention.

1

Pick a hook format

Browse the library above. Each card is one 3-second opening beat with the exact prompt behind it, grouped by the job it does in the first second.

2

Swap in your own line

The prompt opens pre-filled in Studio AI. Change the spoken line to your topic and leave the camera and lighting notes alone — that's what makes it read as filmed.

3

Render and stitch it on

Generate a vertical 3-4 second clip with a spoken hook and matched lip-sync, then drop it onto the front of the video you already have.

The formats

Every hook does
one of five jobs

Pick by intent, not by vibe. The hook that works for a skincare review is not the one that works for a dropshipping ad, and the difference is which of these it's doing.

Impossible shot

Shows something that could not have been filmed at all.

A head held like a bowling ball, a person climbing out of a television. The viewer cannot place what they are looking at, so they watch a second longer to work it out, and that second is the whole game. These only exist because a model rendered them, which is why they are not in anyone else's hook library.

Reaction

Opens on a face already mid-emotion, before any context arrives.

A gasp or a head in hands is legible in a single frame, which is all you get while the thumb is moving. The viewer stays because they want to know what caused it, and you have not told them yet.

Corrective

Tells the viewer they are doing something wrong, then offers the fix.

Being corrected is uncomfortable enough to stop a scroll, and the promise of the fix is what carries attention into the body of the video. It only works if the correction is specific.

Pattern break

Does something visually odd with no explanation attached.

The feed trains people to predict what a frame is about within a moment. An image that refuses to resolve buys you a beat, but you have to earn it back immediately or it reads as a gimmick.

Silent action

No face and no line, just hands doing something absorbing.

Sound-off viewing is the default, so a hook that needs no audio loses nothing. Process shots also imply the payoff is coming, which is its own reason to keep watching.

Every clip in the library above is tagged with the format it belongs to, so you can jump straight to the ones that fit what you're selling.

Back to the library

Frequently asked questions

Everything you need to know about viral video hooks

What is a video hook?+

A video hook is the first three to four seconds of a short-form video — the single beat that decides whether someone keeps watching or swipes past. It is not the intro, the logo, or the setup; it is one line and one shot that earns the next five seconds. On TikTok, Reels, and Shorts the platform measures how many people make it past that opening, and it feeds you more reach when they do. That is why hooks get written and tested separately from the rest of the video.

What is a hook on TikTok?+

The same thing, with a shorter fuse. TikTok's feed drops viewers straight into full-screen playback, so your hook is competing against a thumb that is already moving. The hooks that survive there open mid-action rather than mid-introduction: a raised hand, a sentence already in progress, a box already being torn open. Every clip in the library above is built for that first-second-matters format rather than for a video that gets to warm up.

How long should a video hook be?+

Three to four seconds. Long enough to land one spoken line and one visual beat, short enough that you have not spent your attention budget before the actual content starts. Every prompt on this page is written to that length and set to 9:16 vertical, so what you generate drops straight onto the front of a TikTok, Reel, or Short without trimming.

Are these viral hooks free?+

Yes. Browsing the library and copying any prompt is free and needs no signup. You generate the actual clip in Studio AI, Creative Fabrica's AI video generator, which has a free tier — so testing five hook angles costs you nothing but the time it takes to render them.

How do I use these hooks for Reels and TikTok?+

Pick the hook format that fits what you are selling or saying, take its prompt into Studio AI, and edit the spoken line to your own topic before you generate. The prompt handles the hard part — the handheld phone-camera look, the real room, the uncorrected skin, the single unbroken take — so the only thing you need to write is the sentence. Then stitch the rendered clip onto the front of the video you already have.

What makes a UGC hook actually work?+

It does one job in one second: show something that could not have been filmed, land a reaction, correct something the viewer is doing wrong, set up a formula worth finishing, break the visual pattern, or pull them into an absorbing action. The clips in this library are sorted by which job they do, so you can pick by intent instead of guessing. The common failure is a hook that is technically well shot but does none of them, so it reads as an ad and gets treated like one.

Do I need to write a script?+

Barely. Where a card has its prompt written, that prompt carries the delivery notes and the exact shot description, which is most of what a UGC script is at this length. Swap in your own topic and leave everything else alone: the camera, lighting, and performance direction are what make it read as filmed rather than generated.

Is this an AI hook generator?+

It is a library rather than a generator, and that is deliberate. A hook generator hands you text that still has to be filmed. These are finished video prompts: each one opens pre-filled in Studio AI and renders an actual 3-4 second hook clip with a spoken line, no camera and no creator needed.

How is this different from the AI UGC ad tool?+

The UGC ad tool takes a product photo and builds a full multi-clip ad storyboard around it. This page does one thing: the opening beat, on its own, with nothing to upload. Use this when you already have a video and it is not holding attention; use the ad tool when you need the whole thing.

Which AI video model should I use for hooks?+

One with native audio, because a hook is a spoken line and the mouth has to match it. The prompts open with Veo 3.1 pre-selected in Studio AI for that reason. They also run well in Kling 3.0 and SeeDance 2.0 if you want to compare takes — Studio AI gives you all of them under one plan.

The guide

Viral video hooks, explained

What a hook is, why the first three seconds decide your reach, the formats that keep working, and how to generate one instead of filming it.

What is a video hook?

A video hook is the opening three to four seconds of a short-form video — one line and one shot whose only job is to earn the next five seconds. It is not the intro, the branding, or the setup. On TikTok, Instagram Reels, and YouTube Shorts the feed drops viewers straight into full-screen playback, so a hook video is competing with a thumb that is already in motion. The platforms measure how many people make it past that opening and hand out reach accordingly, which is why experienced creators write and test viral hooks separately from the videos they sit on top of.

The practical consequence: a mediocre video with a strong hook outperforms a strong video with a weak one, every time. Fixing the first three seconds is the cheapest edit available to you.

The five hook formats that keep working

Nearly every hook that survives the scroll is doing one of five jobs in its first second. The library above is sorted by which one, so you can pick by intent rather than by vibe:

  • Impossible shot

    Shows something that could not have been filmed at all. A head held like a bowling ball, a person climbing out of a television. The viewer cannot place what they are looking at, so they watch a second longer to work it out, and that second is the whole game. These only exist because a model rendered them, which is why they are not in anyone else's hook library.

  • Reaction

    Opens on a face already mid-emotion, before any context arrives. A gasp or a head in hands is legible in a single frame, which is all you get while the thumb is moving. The viewer stays because they want to know what caused it, and you have not told them yet.

  • Corrective

    Tells the viewer they are doing something wrong, then offers the fix. Being corrected is uncomfortable enough to stop a scroll, and the promise of the fix is what carries attention into the body of the video. It only works if the correction is specific.

  • Pattern break

    Does something visually odd with no explanation attached. The feed trains people to predict what a frame is about within a moment. An image that refuses to resolve buys you a beat, but you have to earn it back immediately or it reads as a gimmick.

  • Silent action

    No face and no line, just hands doing something absorbing. Sound-off viewing is the default, so a hook that needs no audio loses nothing. Process shots also imply the payoff is coming, which is its own reason to keep watching.

The common failure is a hook that is technically well shot but does none of them. It reads as an ad, so the feed treats it like one.

How long should a hook be?

Three to four seconds. That is enough for one spoken line and one visual beat, and short enough that you have not spent the viewer's patience before the actual content begins. Every prompt on this page is written to that length in vertical 9:16, so the rendered hook clip drops onto the front of a Reel, TikTok, or Short without trimming.

A hook that runs past five seconds is not a hook any more, it is an intro — and intros are the single most common reason short-form videos lose their audience before the point arrives.

How to generate a hook instead of filming one

You do not need a camera, a creator, or a ring light. The whole thing takes about a minute:

  1. Pick the format, not the faceScan the library for the job you need done in second one: reaction, corrective, explainer, pattern break, or silent action. That decision matters more than which creator is on screen.
  2. Swap in your own lineThe prompt opens pre-filled in Studio AI. Change the spoken line to your topic and leave the rest alone: the camera, lighting, and delivery notes are what make it read as filmed rather than generated.
  3. Render it in 9:16Hit generate and you get a short vertical clip, with matched lip-sync where the hook is spoken, ready to sit on the front of the video you already have.
  4. Test five, keep oneThe prompts are free, so run the same format with five different lines and post the one that holds attention. Hook testing is the only reliable way to find out what your audience stops for.

Why these look filmed, not rendered

A hook fails the moment it looks like an ad, and AI video defaults to looking like an ad: clean lighting, smooth skin, a camera on a rig. Every prompt here pushes hard the other way. They specify a front-facing phone lens with its mild barrel distortion, uncorrected skin with visible pores and flyaway hair, a real room with its own clutter, and one unbroken handheld take with the sway that comes from an actual hand. The UGC hooks that convert are the ones that look like somebody grabbed their phone, and that is a prompt-engineering problem, not a model problem.

Hooks vs. full UGC ads

These two jobs get confused constantly. A hook is one beat with nothing to upload — use it when you already have a video and it is not holding attention. A full UGC ad is a multi-clip storyboard built around your actual product, with the hook, problem, product, and benefit beats in sequence. The AI UGC ad tool does that one: you attach a product photo and it builds the whole spot. Most people end up using both — the ad tool for the body, this page for the front three seconds.

Output is vertical 9:16 by default — sized for TikTok, Instagram Reels, and YouTube Shorts, which is where hooks are worth anything in the first place.

Ready to fix your first three seconds? Scroll up, pick the format that fits your niche, and take its prompt into Studio AI.

Your video needs
a better opening

Pick a hook format, swap in your line, and render a 3-second opener in Studio AI. Free, no signup.

Learn how it works

How to Turn Product Photos Into a Video (Free, No Filming)

Use a photo as a reference frame and let AI build a short 3–5s clip around it. The method, the prompt formula, and three prompts you can copy.

Read the guide

Image to Video AI: How to Turn Any Photo Into a Video (Free)

Start from a real photo (especially for faces), the motion-prompt formula, and why SeeDance 2.0 beats Kling, Veo, Grok, and Midjourney.

Read the guide