How to Write a Video Script That Holds Attention
To write a video script that holds attention, open with the most interesting thing you have, give the body a shape the viewer can feel (hook, value, payoff), and write every line to be spoken rather than read. The work is not in sounding clever. It is in earning each next ten seconds and never letting the rhythm flatten out.
Most advice about scripts fixates on the opening seconds and then goes quiet about everything after. That gets the emphasis half right. The opening matters, but the drop-off is gradual and continuous, so a script has to keep re-earning attention all the way through. The numbers below show where viewers actually leave, and the rest of this guide is about writing so they stay.
One thing to clear up first, because it shapes how people write hooks. You may have heard that human attention spans have fallen to eight seconds, shorter than a goldfish. That figure is not real. The BBC traced it to a 2015 Microsoft Canada report that cited an uncorroborated source, and Microsoft's own research never contained the number or the goldfish (BBC News, Mar 2017). What is real is that collective attention is narrowing: one peer-reviewed study found the time a topic stayed in the top 50 trending hashtags fell from 17.5 hours in 2013 to 11.9 hours in 2016, about a 32% drop (Nature Communications, Apr 2019). You are competing for a slice of a shrinking pie, not racing a fish.
What goes in a video script
A video script is the spoken words plus the visual directions that tell a production what to show while those words are said. The classic format is two columns: audio on one side (narration, dialogue, sound), visuals on the other (what is on screen, the shot, any on-screen text). You do not always need the full two-column layout for a short social clip, but you do always need to separate what is heard from what is seen, because they are written differently and someone (or something) downstream has to turn each into a shot.
A useful length rule: spoken English runs around 150 words per minute, so a 60-second video is roughly 150 words of narration, and a 5-minute explainer is around 750. Write to that budget early. It stops you from cramming a four-minute script into a 90-second slot and discovering the problem in the edit.
Start with the hook
The first line is the one line you cannot afford to waste. The defensible version of "the opening matters" comes straight from the platforms: YouTube's own Intro metric measures who is still watching after 30 seconds and tells creators to surface compelling content earlier (YouTube Help), and Google found that the first five seconds of a skippable ad are what earn the rest of the view (Think with Google, Jun 2015). So open with the payoff, the tension, or the single most surprising thing you are going to say. Do not warm up. Do not introduce yourself before you have given them a reason to care who you are.
Do
- Open on the payoff or the boldest claim you can back up.
- Start with a concrete question the viewer already has.
- Drop them into a moment that is already in motion.
- Make a promise in the first line that the rest of the script keeps.
Don't
- Spend the first 15 seconds saying hello and setting up.
- Explain what the video is about before showing it.
- Bury the interesting line three paragraphs in.
- Ask viewers to "stick around to the end" before you have earned a second.
A trick that works: write the hook last. Draft the whole script, find the most interesting sentence buried in the middle, and move it to the top. The thing you were saving for the reveal is usually the thing that should have opened the video.
Give it a structure the viewer can feel
A reliable shape for most short and mid-length video is hook, value, payoff. You promise something, you deliver the substance they came for, and you close with a result or a turn that makes the whole thing feel complete. The structure matters because attention leaks at every seam where the video stops giving and starts coasting, not only at the start. Wistia's data shows engagement holds reasonably flat from one to five minutes and then falls off past five (Wistia 2026 State of Video, Jul 2025), which is the practical case for sections: each new section is a chance to re-hook instead of drift.
Hook
The promise or the tension, up front. Tell the viewer what they get and why it is worth the next minute, in one or two lines, then move.
Value
The steps, the story, or the argument that pays off the hook. Break it into beats so each one delivers something new instead of stretching one idea thin.
Payoff
The result, the lesson, or the twist that lands it. Close the loop you opened, then a single clear call to action. One ask, not five.
For anything past a few minutes, treat each section like its own miniature of that shape: a small hook to open it, the substance, a small payoff that hands off to the next section. A 12-minute explainer that re-hooks every two minutes holds far better than one that front-loads a great opening and then lectures.
Write for the ear, not the page
Scripts are spoken, so they have to survive being read aloud. Use short sentences. Cut clauses. Read every line out loud and delete anything you stumble over, because if you trip on it, the AI voiceover will too. Contractions and plain words sound like a person; dense, formal phrasing sounds like a press release being recited. The single most useful editing pass on any script is reading the whole thing aloud at the speed you mean it to play.
Compare the two below. Both say the same thing. Only one of them sounds like a human talking.
Reads like a memo
"In order to maximize retention, it is essential that creators prioritize the establishment of narrative tension within the initial segment of the content."
Sounds like a person
"Want people to stay? Give them a reason to in the first line. Then keep your promise."
Control the pacing
Attention drops when a video settles into one rhythm and stays there. The fix is variation, in the writing and in what it implies on screen. Follow a long explanation with a short, punchy line. Change what is on screen often, and let the script imply those cuts instead of describing one static frame for 40 seconds. A working rule: if a stretch of script has not given the viewer something new (a fact, a turn, a visual change) in about 15 seconds, tighten it or cut it.
Format the script so AI can build it
If you are turning the script into video with AI, the format does real work. The cleanest path is to write the script so a workspace can read it into scenes, then into shots, then into footage and narration. Adoption of this workflow is no longer niche: the share of brands using AI in video production jumped from 18% in 2024 to 41% in 2025, more than doubling in a year (Wistia 2025 State of Video, Mar 2025). The script is the input that decides whether the output is coherent or mush.
The lever most writers miss is the content type: how the script is read into scenes. A narration script, a presenter script, an explainer, and a commercial all break down differently, so telling the workspace which one you are writing changes the entire shot breakdown. In Fawna, you pick that content type up front, and the script you paste is parsed accordingly into a shot-by-shot AI storyboard, with a consistent voiceover read off the same lines. To make that hand-off clean, write with a few habits in mind.
- Mark who is speaking. Narrator, a named character, on-screen text. Each maps to a different voice or a different element.
- Keep visual directions on their own lines. What is heard and what is seen are different jobs; do not blend them into one paragraph.
- Let structure mirror the video. One scene per beat, in order. The clearer the script's shape, the closer the generated board is to what you pictured.
Script input is paste today (PDF and image upload are coming), so the practical move is to draft anywhere, then paste a clean, sectioned script in. A muddy script produces a muddy board no matter how good the model is. A clear one means the bottleneck moves back to ideas, where it belongs. If you want to see the full hand-off, the guide on what an AI storyboard is explains the layer in between.
Frequently asked questions
How long should a video script be?
Match the script to the runtime. Spoken English runs about 150 words per minute, so a one-minute video is roughly 150 words of narration and a five-minute explainer around 750. Write to that budget from the start, because it is far easier to trim a tight script than to find time you do not have in the edit.
How do you start a video script?
Open with the single most interesting thing you have: the payoff, a sharp question, or a moment already in motion. Skip the greeting and the setup. A common trick is to draft the whole script, find the best line buried in the middle, and move it to the very top.
What is the standard video script format?
The classic format is two columns: audio on one side (narration, dialogue, sound) and visuals on the other (what is on screen, the shot, any text). Short social clips can be simpler, but always separate what is heard from what is seen, because each becomes a different part of the finished video.
Should I write a script word for word or use an outline?
It depends on delivery. Word for word suits narration, ads, and anything voiced by AI, where exact wording and pacing matter. A bullet outline suits a confident on-camera presenter who sounds stiff reading verbatim. If you plan to generate the video from the script, write it word for word so the breakdown is precise.
What is the difference between a script and a storyboard?
The script is the words and directions in text; the storyboard is the visual version, panel by panel, showing how each moment looks on screen. The script comes first and feeds the storyboard. In an AI workflow, a clean script is read into scenes and then drawn out as a storyboard automatically.
How do you write the call to action at the end?
Make one ask, not several, and tie it to the payoff you just delivered. After a how-to, point to the next step; after a story, invite the obvious follow-on. Keep it short and specific ("try this on your next video"), and place it right after the close, while attention is still there.
