Script Voice Over Example for a 60-Second Explainer

Below is a finished script for a sixty-second product explainer, formatted the way a narrator, an editor and a text-to-speech engine all need to read it. Say the right column out loud at a conversational pace. It lands near a minute with silence to spare.

The narration above runs about 130 words, not 180. That gap is deliberate: it is the breathing room that keeps a read from sounding rushed when a narrator takes a real breath at each // mark.

Notice what the right column does not contain. No stacked clauses, no sentence that needs a comma to survive, no number the narrator has to interpret on the fly. Each beat carries one idea and hands off to the next.

The left column is not decoration either. Pairing each line of narration with the shot it sits under is what stops you from writing forty words over a three-second cutaway.

Time needed: 5 minutes. What you need: your draft, a word counter, a stopwatch.

Count words before you write sentences. A comfortable explainer read sits somewhere between two and two and a half words per second, which puts a one-minute slot at roughly 120 to 150 words once you subtract pauses. Treat 150 as a ceiling you rarely touch, not a target.

People speak far faster than they can be comfortably followed. Research by Romero-Fresco on BBC programmes found speakers reach 245 WPM in interviews, well above comfortable reading and narration speeds. That number is a warning, not a model — an explainer read at interview speed is unlistenable.

Reading limits set elsewhere in the industry point the same direction. Netflix’s English (USA) Timed Text Style Guide caps subtitle reading speed at 20 characters per second for adult programs, which is a useful outer edge when your narration also has to be mirrored on screen.

Write your draft long, then cut to budget. A first pass at 200 words is normal; the cut from 200 to 140 is where the script gets good, because you delete qualifiers rather than ideas.

Time needed: 10 minutes. What you need: the four-beat grid below and your word budget.

Divide the budget before you divide the time. The grid that holds up across product explainers, course intros and internal announcements looks like this:

Short videos earn their length. Wistia’s 2025 State of Video report found videos under 1 minute reached an 82% engagement rate, versus 58% for videos running 5–30 minutes. Wistia’s 2025 State of Video findings are drawn from more than 14 million videos, 100,000 businesses and a survey of 1,300 professionals.

The first beat is where most scripts leak time. Opening with a company name, a greeting or a definition burns four seconds that the hook needed.

Keep one call to action, in the final beat, stated once. Two asks in sixty seconds produce zero.

Time needed: 20 minutes. What you need: a two-column table in any doc or sheet, plus your shot list.

A single-column script hides timing problems. Two columns force every narration line to answer a question: what is the viewer looking at while they hear this? If the left cell is empty, the line is probably filler.

Formatting rules that keep both columns usable:

• Left column: one shot per row, described as an action, not an adjective.
• Right column: spoken words only — no stage directions mixed into the narration.
• A third narrow column for timecodes if your editor works from timecode rather than row order.
• On-screen text written out exactly as it will appear, so it never duplicates the narration word for word.

On-screen lines have hard ceilings worth borrowing. According to 3Play Media, Netflix limits captions to 42 characters per line, which constrains how much narration can be mirrored on screen at once. Netflix also requires every subtitle event to remain on screen for a minimum of five-sixths of a second, a useful floor when you are timing short on-screen lines against narration.

Ship this two-column file to the voice talent, the editor and the localization lead at the same time. One artifact, three readers, no version drift.

Time needed: 10 minutes. What you need: a consistent mark-up legend at the top of the script.

Mark-up is not decoration — it is the difference between a read that matches your timing math and one that drifts. Pick a legend, put it at the top of the document, and never change it mid-project.

Spell out anything a stranger could misread: product names, acronyms, URLs, abbreviations, and any number you decided to keep. A narrator will ask; a synthetic voice will simply guess and move on.

Synthetic voices follow punctuation more literally than people do, so commas and periods become timing instructions. Platforms such as AI Studios offer more than 1,000 AI voices and text-to-speech in 150+ languages and accents on paid plans, which means the same marked-up script can be auditioned in several reads before you commit to one.

Time needed: 15 minutes per pass, two or three passes. What you need: a stopwatch, a phone recorder, a quiet room.

Estimating is not timing. Record yourself reading the right column at the pace you actually want on screen, then look at the clock — the gap between your estimate and the recording is usually five to eight seconds, always in the wrong direction.

Cut in this order: adverbs, then qualifying clauses, then the second example, then whole sentences. Never cut the pauses first, because pauses are what make a 130-word script sound unhurried.

Leave headroom if the video will be dubbed. Most languages expand when translated from English, and the ceilings differ by market: Netflix’s Greek Timed Text Style Guide sets lower reading-speed ceilings than English, allowing up to 17 characters per second for adult programs and up to 13 for children’s programs. For English subtitle templates, Netflix allows up to 17 characters per second on adult programs and up to 15 characters per second on children’s programs.

A practical habit: bring the English read in near fifty-five seconds rather than fifty-nine. That four-second cushion absorbs expansion without forcing your dub into a sprint.

Run these ten checks on the final file. Each one catches a problem that is cheap to fix now and expensive to fix after the session.

1. Spoken word count sits under 150 for the full minute.
2. The hook lands inside the first ten seconds, with no greeting or company intro ahead of it.
3. Every narration row has a paired visual in the left column.
4. No sentence carries more than one idea or more than one comma.
5. Acronyms, product names and numbers all carry a [say: ...] guide.
6. Pause marks sit at scene changes, never inside a phrase.
7. On-screen text lines stay short enough to read comfortably and never duplicate the narration verbatim.
8. Timing comes from a recorded read-aloud, not an estimate.
9. The English read finishes a few seconds short, leaving room for translated versions to expand.
10. One call to action appears, once, in the final beat.

If three or more boxes stay unticked, do another cut pass before booking anyone. Scripts fail in the booth for reasons that were visible on the page.

The failure you will meet most often: the script reads fine silently, runs eight seconds long out loud, and gets rescued by speeding up the narrator. The read turns breathless, the pauses vanish, and the on-screen text now flies past faster than anyone can follow. Fix it by cutting words, never by raising the pace — speaking faster is how narration drifts toward the 245 WPM that Romero-Fresco measured in BBC interviews, which no viewer can comfortably follow.

Four more that show up repeatedly:

• Stacked subordinate clauses. A human narrator can hold them together with intonation; a synthetic voice flattens them. Split into two sentences.
• The late call to action. Anything introduced after the fifty-five second mark gets clipped by the outro. Move it into the 0:45–1:00 beat.
• Narration that reads the screen text aloud. The viewer processes one or the other, not both. Let the text carry the label and the voice carry the reason.
• No pronunciation pass. One unfamiliar product name triggers a retake, and retakes cost more than the ten minutes the pass would have taken.

Rewrite the script after the first recorded read, not before. The page lies about timing; the stopwatch does not.