Writing narration that sounds right out loud

Narration is not copy. Copy is read at the reader's pace; narration is heard once, at the speaker's. A few habits make the difference between a voice that sounds natural and one that sounds like it is reading a brochure.

Short sentences

One idea per sentence. A full stop gives the voice a real pause and gives the viewer time to look at the screen. If a sentence needs a semicolon, it is two sentences.

Commas are breaths

Put commas where a person would breathe. The voice treats them as short pauses, and captions break on them too, which keeps each caption line readable.

Write numbers the way they are said

The voice reads what you write. Use pronunciation tokens so captions stay tidy while the voice says it properly:

[₦55|fifty five naira] per transfer.
Visit [ai.twelveai.app|ai dot twelveai dot app].
Set up [2FA|two factor] in a minute.

Budget your words

At a natural pace, narration runs at roughly 2.6 words per second. A 15 second video is about 35 words. A 45 second video is about 110. Write to the budget, then let the assistant's Fit to length tighten anything that runs over.

Read it aloud once

Before you render, press Hear this line on the scenes that matter most. If a sentence makes you stumble, it will make the voice stumble too.

No exclamation marks

They rarely improve a voice reading and they almost always cheapen the caption. Let the visuals carry the energy.