How to write a video script

Writing for video is writing for the ear, with a picture running alongside it. Two consequences follow, and almost every bad script ignores at least one: sentences must be sayable in one breath, and you must never describe in words what the viewer can already see.

Pace: roughly 140 to 160 words per minute of finished video. A ten-minute video is about 1,500 words. Scripts almost always run long.

The first fifteen seconds

More viewers leave in the opening fifteen seconds than at any other point, and no amount of quality later recovers them. Do not open with a channel introduction, a logo animation, or "hey guys, welcome back".

Open with one of:

• The payoff, stated. "By the end of this you'll be able to read any nutrition label in about four seconds."
• The problem, sharply. "Your sourdough is dense because of something you're doing on day two, not day five."
• A specific, surprising fact.
• An open loop. A question you visibly will answer.

Then a short line on why you can be trusted on it — a credential, an experience, a number — and straight into the substance. Housekeeping goes later or nowhere.

Structure

Hook (0:00–0:15). Above.

Promise and roadmap (15–30s). What they will get. Keep it short; a long contents list is itself a reason to leave.

Body, in segments. One idea each, three to six of them. Each segment: the point, the evidence or demonstration, the so-what. Signpost transitions out loud — "that handles the base; the second thing is temperature" — because a viewer who looked away needs a way back in.

Re-hook every couple of minutes. A question, a change of location or format, a preview of what is coming. Attention decays and has to be re-earned.

Payoff. Deliver what you promised, explicitly.

One call to action. Only one, and place it after value has been delivered rather than at the start.

The two-column format

For anything with visuals, write in two columns: audio on the left, visual on the right.

This forces you to plan what is on screen and stops you writing narration that describes the picture. It is also the format an editor can actually work from.

For talking-head video with no b-roll, a single column is fine — but mark your cuts.

Writing for the ear

Short sentences. One idea each. If you run out of breath reading it aloud, split it.

Contractions. "You'll", "it's", "don't". Written-out forms sound stilted spoken.

Second person. "You" throughout.

Say numbers as you would speak them. "About a third", not "33.4%".

No subordinate clauses stacked up. A listener cannot re-read.

Read every line aloud as you write. This is not a final check; it is the method. Anything you stumble over, rewrite.

Common mistakes

A long intro before any value. Narrating what is visible on screen. Sentences written for the page. Cramming three videos' worth of content into one. Asking for the subscribe in the first ten seconds. No signposting, so the video feels like one undifferentiated block. Ignoring the runtime until the edit, then cutting the best material to fit.

Practical tips

• Write the hook last, once you know what the video actually delivers.
• Time the read-aloud rather than trusting the word count; your pace is personal.
• Mark breath points in the script with a slash if you struggle with delivery.
• Cut 20% after the first read-aloud. It will still be too long.
• Keep a separate file of the b-roll you need; it becomes your shot list.