Voice Over Script for a 30 Second Ad: Word Count & Timing

How many words does a voice over script actually need for a 30 second ad? Most people guess too high, record it, and find the read runs 40 seconds. Then they speed up the audio and it sounds like a chipmunk auctioneer.

The fix is simple arithmetic plus one habit: read the thing aloud with a timer. Here's how to do both, and how to structure the words you do have.

How many words fit in 30 seconds

A natural conversational read runs at about 2.5 words per second, according to Killerspots' radio script timer guide. That gives these rough ranges:

• 15 seconds: 35 to 40 words
• 30 seconds: 70 to 80 words
• 60 seconds: 140 to 160 words

The same page notes that energetic reads can reach about 3 words per second, while warm or contemplative ones slow to about 2.2. It also suggests 65 to 70 words for a 30 second spot with two voices, because the hand-offs take time.

So your target is about 65 to 75 words, not 90. If your ad has music, sound effects or on-screen moments that need to breathe, aim for the low end.

Why not just speed the read up?

Because there's a limit. The same guide says a voice actor can compress a read by "maybe 10 to 15 percent" before quality drops noticeably. Cutting words at the writing stage is much safer than squeezing audio afterwards. AI voices behave similarly: crank the speed and the delivery gets clipped and flat.

What eats your word budget

• Long brand names and unfamiliar product terms
• Phone numbers, URLs and prices (they're slow to say and easy to garble)
• Pauses, which you want to keep

If your script includes a web address, count it as more than its word length suggests. Better still, say "link in bio" or put the URL on screen and keep it out of the voice track.

A beat-by-beat structure for 30 seconds

Think in beats, not paragraphs. This split works for most product and service ads:

1. Hook (0 to 3 seconds, about 8 words). A problem, a question or a surprising claim.
2. Problem or context (3 to 10 seconds, about 18 words). One specific pain, not three.
3. Product as the answer (10 to 22 seconds, about 28 words). What it is and the one thing it does well.
4. Proof or detail (22 to 26 seconds, about 10 words). A concrete detail, only if you can honestly back it up.
5. Call to action (26 to 30 seconds, about 8 words). One action.

That adds up to roughly 72 words. Treat the numbers as a budget, not a rule.

The hook matters most. TikTok's creative best practices page says to prioritize your hook in the first 6 seconds and to introduce your content proposition in the first 3 seconds for better recall and awareness. That page doesn't say anything about voice-over specifically, so take it as guidance on the opening of the ad, not proof that any particular script style wins.

A worked example

Say you're selling a reusable coffee cup. Here's a script at about 70 words:

Your coffee's cold by ten. Again. (hook)

Most travel cups leak, or they keep heat for an hour and give up. (problem)

The Ember cup is double-walled, fits any cupholder, and the lid actually seals. Tip it upside down in your bag. Nothing. (answer and proof)

Still hot at lunch. (payoff)

Get yours at Ember dot com, and make the first cup last. (CTA)

Notice the short lines. They give the editor cut points and give the voice room to land the key phrase. If you want the product to sound like a person rather than a brochure, write the way you'd explain it to a friend.

Write for the ear, not the page

A voice over script is spoken, so a few rules differ from normal copy.

• Use contractions. "It's" reads more naturally than "it is" in almost any ad.
• Keep sentences short. One idea per sentence. If you run out of breath reading it, so will the voice.
• Vary the length. Killerspots recommends mixing short sentences for key lines with longer ones for the middle, so the read doesn't sound monotone.
• Avoid tongue-twisters. Repeated "s" or "sh" sounds across a line are hard for people and models.
• Spell out what's hard to say. Write "twenty twenty-six" or "fifteen dollars" if the numerals might be misread.

Match words to picture

The voice shouldn't describe what we can already see. If the video shows a hand tipping the cup upside down, the voice can say "Nothing." and let the picture do the rest. Talk about what the visuals can't show: the benefit, the feeling, the reason.

Time it properly

Word counts get you close. A timer gets you there.

The Killerspots guide recommends reading the script aloud at performance pace three times and averaging the results, since silent scanning gives unreliable estimates. It's a good habit. Do it before you generate a single frame of video.

Then trim:

1. If the average is over 30 seconds, cut words, not pauses.
2. If it's 27 to 29 seconds, you're in good shape. That leaves a beat for a logo or end card.
3. If it's under 25, add a pause or a line of proof rather than padding.

Fixing a script that an AI voice reads badly

If you're generating the voice with text to speech, you'll find that written punctuation becomes direction. ElevenLabs' best practices documentation is a useful primary source here, though what it describes applies to ElevenLabs' own models and other voice tools may behave differently. Per that page:

• On Eleven v4, punctuation "significantly affects delivery." Ellipses add pauses and weight, and capitalizing a word increases its emphasis.
• Eleven v4 and v3 don't support SSML break tags, so pauses come from ellipses and text structure. Older models support <break time="x.xs" /> for pauses up to 3 seconds, and overusing them can cause instability.
• Models can mispronounce phone numbers, currencies, dates and abbreviations. The docs suggest expanding abbreviations to their full spoken forms before generating.
• Speed settings range from 0.7 to 1.2 with 1.0 as the default. Pacing also depends on the voice itself.
• For complex sequences, the docs recommend generating segments separately and layering them in audio software.

That last point is practical for a 30 second ad. Generate the hook, the middle and the CTA as three separate takes. If one line comes out wrong, you redo ten words, not seventy. Results vary from run to run, so expect to generate a line a few times and pick the best.

Whatever tool you use, listen for the same things: stressed words in the wrong place, numbers read oddly, and a CTA that trails off. Rewrite the line rather than fighting the voice. "Get yours at Ember dot com" can read better than "Get yours at Ember.com", for instance. Test it, because it depends on the voice.

Where Mediafy Studio fits

You can use it at a few points, none of them required.

• Drafting. Mediafy Chat can help you brainstorm hooks and tighten a draft. Paste in your script and ask for a version at 70 words. Always run the timer on whatever comes back, because word counts from any assistant are approximate.
• Making the audio. In the Studio you can pick Audio, choose a model and write a prompt. If you're building a whole ad, Flows can chain an image step, a video step and a voice-over step in one canvas.
• Talking-head formats. UGC Factory writes a short script for a character and product, makes the selfie-style image, lip-syncs it into a talking video and adds subtitles. You'd still review that script against the word budget above.

If you want speech generated with the video itself, read Veo 3.1 Native Audio: How to Prompt Sound and Speech. Spoken lines inside a video clip follow different rules than a separate voice track. And since a 30 second ad is usually vertical, check Aspect Ratio for Social Media before you plan the visuals.

Common mistakes

• Writing 100 words and hoping. Do the math first.
• Three messages in one ad. One problem, one product, one action.
• Reading the script silently. It always feels shorter in your head.
• Putting a URL or phone number in the voice track. Show it on screen instead.
• Not leaving an ending beat. Give the last line a second of room.
• Testing only one version. Write two or three hooks and compare them.

A quick checklist

1. Aim for 65 to 75 words.
2. Split it into five beats.
3. Write short, spoken sentences with contractions.
4. Read it aloud three times with a timer and average.
5. Cut to land at 27 to 29 seconds.
6. Generate the voice in short segments and replace any that sound off.

Want to turn a finished script into a voiced video? You can try it at Mediafy Studio with a Google account.

Sources

• Killerspots, "Radio script timer: how long is a 30 second script": https://killerspots.com/blog/radio-script-timer-how-long-is-a-30-second-script
• ElevenLabs, "Best practices" (text to speech documentation): https://elevenlabs.io/docs/overview/capabilities/text-to-speech/best-practices
• TikTok for Business, "Creative best practices for performance ads": https://ads.tiktok.com/help/article/creative-best-practices?lang=en