To write a tutorial script for AI narration, structure it for attention and for the screen. Hook the viewer in the first 15 seconds and promise the result. Give one action per sentence, using the exact on-screen labels, with exact pauses for clicks. Change the pattern every 2 to 3 minutes and end with one clear call to action.
The general rules of writing for the ear are in our primer, How to write a script for AI voices. This post covers what's different about tutorials.
What structure works for a tutorial script?
| Part | Length | Job |
|---|---|---|
| Hook | First 15 seconds | Show or state the result the viewer wants |
| Promise | One or two sentences | Say what they'll be able to do by the end |
| Steps | Most of the video | One action per sentence, in screen order |
| Pattern interrupts | Every 2 to 3 minutes | Reset attention: a result, a mistake to avoid, a question |
| Call to action | Last 15 to 30 seconds | One next step |
How do I write the hook?
Lead with the outcome, not the introduction. "In the next five minutes, you'll turn a blog post into a narrated video" beats "Hi, welcome to the channel, today we're going to talk about...". Show the finished result on screen while the hook plays if you can. More hook techniques: The 5-second hook.
Then make the promise concrete: what they'll have, and roughly how long it takes.
How do I write steps that match the screen?
Give each action its own sentence. "Click Settings. Then choose Pronunciation." is easier to follow than one sentence with both. Say things in the order the viewer sees them, naming where to look before what to click: "In the top right, click Export."
Use the exact labels. If the button says Create backup, don't say "make a backup", because viewers search the screen for the words you say. After an action, say the result: "A dialog opens." tells the viewer they did it right.
Write shortcuts and paths as spoken: "Command S", not "Cmd+S"; "Settings, then Pronunciation", not "Settings > Pronunciation".
How do I leave time for screen actions?
Narration that runs straight over a menu opening loses the viewer. Add an exact silence where the screen needs time:
- In Vois Studio, type
/, choose Pause and pick 0.3 to 3 seconds. The amber pill holds that silence in the audio. - Use short holds for clicks and longer ones for loading or a result appearing.
- Don't rely on commas or full stops for this. Punctuation shapes rhythm, but it doesn't set a length.
Details: Pause nodes for precise silence.
Where do on-screen cues and notes go?
Keep them in the script but out of the spoken text. In Vois Studio, Heading, Title, Section and Note blocks from the / menu are never spoken. Use them for chapter names, "zoom in here" cues and timing reminders, so the person editing the video can follow the same document.
What is a pattern interrupt?
In a long tutorial, attention drops when every minute sounds the same. Every 2 to 3 minutes, change something:
- Show the result so far.
- Name a common mistake and how to avoid it.
- Ask the viewer a question: "Does your menu look different? Here's why."
- Start a new section with a short summary of what's next.
How do I handle product names, numbers and versions?
Product and feature names that a voice gets wrong belong in a pronunciation dictionary, so every tutorial says them the same way: Pronunciation dictionary guide.
Write version numbers, prices and settings as spoken: "version two point four", "twelve dollars", "forty-eight kilohertz". See How to write numbers, dates and acronyms for AI voices. Say URLs simply ("the link is in the description") or spell them out as spoken.
A sample tutorial script
[Narrator]: Want your scripts to say every product name right? In two minutes, you'll fix it for good.
By the end, any name you add will be pronounced the same way in every project.
First, open Settings. [pause=500] Then choose Pronunciation.
Click Add. [pause=500] In Word, type the name as it's written. In Pronunciation, type how it should sound. Then click Add again.
That's it. The rule is used the next time you generate.
Next, let's load a whole glossary at once.
How long will it run?
At about 150 words a minute in English, the sample above is under a minute of speech, plus its pauses. Use that rate only to plan. Generate the script in the voice you'll use, measure the audio, and cut the video to the real narration: How long is a script read aloud?
How do I end?
With one call to action, said plainly: "Try it with your own script, and the next video covers exports." One ask works better than a list. Before you generate, run the AI narration script checklist.
Related
- How to write a script for AI voices
- AI narration script checklist
- How long is a script read aloud?
- Turn release notes into product tutorial audio
- The 5-second hook: voice tricks for YouTube retention
Pick one task you explain often, write it as a two-minute script tonight, and generate it before you record a single frame.
The Vois Team