Writing a 30-second video ad script

Before you write anything down

• Name the one problem. Not the category, not the mission. The specific annoyance a person recognizes in themselves. If you cannot say it in a sentence, you do not have an ad yet.
• Name the one action. Download, sign up, buy, book. One. A script that asks for two things gets neither.
• Know who is talking. A founder, a customer, a narrator, or a presenter reading to camera. It changes the pronouns and the tone before you write a word.
• Have the product name ready to use early. Not at the end as a reveal. Most people who see a short ad will never watch it to the end, so a name that only appears at second twenty-eight reaches almost nobody.
• Know what the viewer will be looking at. A script written blind against footage that does not exist yet will fight the visuals later.

The word-count math

Work backwards from the clock. At roughly 2.5 words per second, thirty seconds is about 75 words of speech. Then subtract: most short ads end on a card or a logo shot with no dialogue over it, which costs you two or three seconds. So the spoken script is closer to 70 words, and if you write 110 you will either be rushed on the read or you will run long.

A workable split for those seventy-odd words is roughly a fifth on the hook, a bit over half on the demonstration, and the rest on the call to action. In seconds that is something like five up front, sixteen in the middle, and six to close. These are not laws — they are a starting shape, and the point of having one is that it stops you from spending fourteen seconds on a preamble.

• Hook — about 15 words, 5 to 6 seconds. The problem, stated as a problem. This is the only part most viewers hear.
• Demonstration — about 40 words, 15 to 17 seconds. What it does, shown while it is said. One capability, not a feature list.
• Call to action — about 15 words, 5 to 6 seconds. What to do, where, and why now.
• End card — 2 to 3 seconds, no dialogue. Name, logo, and the ask in writing for the people who watched with sound off.

Step by step: writing the seventy words

1. 01Write the last line first. The exact ask, in the words you want ringing at the end. Everything before it now has a destination.
2. 02Write the problem in one sentence, using the words your customer would use, not the words your product page uses. "Editing a video takes all evening" beats "streamlining post-production workflows."
3. 03Draft the hook from that sentence. Three lines at most. Lead with the problem, not with the product, not with a greeting, and never with "Hi, I'm...".
4. 04List every feature you want to mention, then delete all but one. The one that survives is the one the demonstration shows. Keep the list — it is the raw material for the next ad.
5. 05Say the product name by the second line. Early enough that a viewer who leaves at eight seconds still knows what they saw.
6. 06Write the demonstration as things happening, not as adjectives. "You paste the link and it writes the script" is a demonstration. "Powerful AI-driven creation" is not.
7. 07Break it into short lines, one idea each. A line should be sayable in a single breath. This is also what makes captions readable, since captions break on natural pauses.
8. 08Read it out loud with a stopwatch. Out loud, at ad pace, not in your head. The stopwatch is the only honest editor you have.
9. 09Cut to length by removing whole ideas, not by squeezing words out of every sentence. A script trimmed word by word ends up dense and unspeakable; a script trimmed idea by idea stays clean.
10. 10Read the first three seconds with the sound off. If the opening line does not work as text on screen, rewrite it, because a large share of viewers will only ever read it.

A worked example of the shape

It helps to see the structure fixed in place. Vidpop, an iPhone app that makes short vertical video ads, writes every script to exactly the same skeleton: 11 lines, split 3 hook, 5 demonstration, 3 call to action. Its finished reels land at about 25 to 35 seconds.

That is a useful reference point whether or not you use the tool, because it makes the arithmetic concrete. Eleven lines across roughly thirty seconds is under three seconds a line, which forces each line to be short — one clause, one idea, no subordinate clauses stacked up. Three lines is enough to state a problem and no more. Five lines is enough to show one thing properly. Three lines is enough to ask once and say where.

You can borrow the skeleton directly: write eleven short lines, three of them setting up the problem, five of them showing the thing working, three of them asking. If any line will not fit in one breath, it is two lines or it is cut. Count the total words at the end — if you are far past eighty, an idea needs to go, not a word.

Rules of craft that survive contact with a real read

• Write for the ear, not the page. Contractions, short sentences, no parentheses, no semicolons. If you would not say it, do not write it.
• One idea per line. The moment a line contains "and also", it is two lines.
• Use the second person. "You" appearing more often than "we" is a reliable signal that a script is about the viewer rather than the company.
• Cut the adjectives first. "Fast, beautiful, powerful" costs three seconds and proves nothing. The demonstration proves it or nothing does.
• Avoid numbers you have to explain. A statistic that needs context is a statistic that eats five seconds.
• Do not open with your logo. The first two seconds are the most valuable ones you will ever have and a logo spends them on someone who does not know you yet.
• End on a verb. The last thing heard should be an instruction, not a slogan.

What goes wrong

• It runs long. Almost always because the script has two ideas in the middle. Cut one; save it for the next ad.
• The hook is a feature. "Now with AI captions" is not a hook. The problem the captions solve is the hook.
• Nobody knows what the product is called. The name is buried at the end. Move it to line two.
• It reads well and sounds terrible. Written-English rhythm does not survive being spoken. Read every draft aloud; the sentences that trip you are the ones to rewrite.
• There is no ask. The script describes and then stops. Ending on a description leaves the viewer with nothing to do.
• It is all "we". "We built", "our platform", "our mission". Swap the subject of each sentence to the viewer and see how many sentences survive.
• The captions do not match the audio. If your on-screen text comes from the script draft rather than from what was actually said, the two will drift apart on any ad-lib or rephrase. Vidpop sidesteps this by taking caption text from a transcript of the rendered voiceover rather than from the script file, so the words on screen match the words spoken.

Frequently asked questions

How many words is a 30-second video ad script?

About 70 to 80 words. Ad reads run at roughly 2.5 words per second, and a few seconds usually go to a silent end card, so 70 is a safer target than 80. Vidpop's generated scripts sit in the same territory — 11 short lines producing a reel of about 25 to 35 seconds.

How long should the hook be in a 30-second ad?

Five to six seconds, or roughly 15 words — about three short lines. That is the structure Vidpop uses for every script it writes, with three lines of hook, five of demonstration and three of call to action, and it works because most viewers never hear anything past the hook.

Should the product name go in the first line?

By the second line at the latest. Short ads lose most of their audience early, so a name held back for a reveal at the end reaches almost nobody. Vidpop's own scripts introduce the product inside the opening three lines for the same reason.

Can AI write a video ad script that is actually usable?

It can write a solid first draft against a fixed structure, which is most of the value. Vidpop writes an 11-line script from your app or product details and lets you rewrite it in its edit chat afterward, so the AI handles the shape and you handle the voice.

Can I set how long the finished ad is?

In some tools, no. Vidpop derives reel length from the script — speaking time plus a short end card — landing at roughly 25 to 35 seconds, and there is no control for setting a duration directly. If you want it shorter, the script has to get shorter.

Do on-screen captions have to match the script exactly?

They have to match what is spoken, which is not the same thing. Vidpop generates caption text from a transcript of the rendered voiceover rather than from the script draft, so any difference between the written line and the delivered line does not end up as text that contradicts the audio.

How do I write a 30-second ad script in Arabic?

Write it in Modern Standard Arabic if it needs to travel across the region, and count spoken time rather than translating word for word — a direct translation of an English script frequently runs long. Vidpop writes its Arabic scripts in Modern Standard Arabic and offers regionally specific voices including Saudi, Kuwaiti, Omani and Egyptian.

What is the ideal structure for a short video ad?

Hook, demonstration, call to action, in that order and roughly in a 1-3-1 ratio of time. It is the structure Vidpop bakes into every script it writes, and its value is mostly negative — it stops you spending fifteen of your thirty seconds on preamble.