To narrate a course module with AI, you need ten short scripts, one voice and about 380 points for a 10-slide module. You get back ten audio files, one per slide, ready to drop onto the slides in PowerPoint or an authoring tool. What decides whether it sounds like a course or a robot reading a deck is the script, not the voice.
This is the whole workflow for one module: how long each script should be, how to pick a voice you can keep for months, what to do about terms the voice gets wrong, and the points math with a real set of slide lengths.
Add narration to slides: write one script per slide
Do not paste your slide text into the generator. Slide text is written to be skimmed, in fragments, and a voice reading fragments sounds like a voice reading a list. Write a separate script for each slide, the way you'd say it to one person across a desk.
Aim for 30 to 60 seconds per slide. In the app a point buys about 17 characters, and a point is about a second of audio, so a 45-second slide is roughly 765 characters, or somewhere near 125 words. Longer than a minute and learners drift. If a slide needs two minutes, it is two slides.
Slide text read aloud: "Three benefits: faster onboarding, fewer errors, consistent training."
Script written for the ear: "Why bother? Because new hires get up to speed faster, they make fewer mistakes, and everyone gets the same training."
Keep one idea per slide script, end each on a full stop, and avoid sentences you'd have to take a breath in the middle of. Numbers are easier to hear when you round them: "about two hundred" beats "203" unless the exact figure matters.
AI voiceover for an online course: pick the voice once
Choose the narrator before you write the other nine scripts, because a voice that suits a lively intro may wear on you across eight modules. For course material, look for a voice described as articulate, composed or steady. You want one that disappears behind the content. The three below are worth auditioning on your own first slide, since a voice that sounds right on a demo line can behave differently on your wording.
Articulate Male Explainer Voice
Technical modules and software tutorials
- Neutral American
- Articulate, precise
Articulate Female Professional Voice
Corporate training and compliance modules
- Neutral American
- Articulate, professional
Long modules where the voice should stay out of the way
- Neutral American
- Composed, steady
Audition with real text. Paste your longest, most technical slide script and listen to the whole thing, including the sentence with the hardest term. A voice that handles that one will handle the rest. Browse the full voice library if none of these fit.
Use your own voice for the course
If learners already know your voice from live sessions, clone it from a 15 to 30 second sample (30 seconds is the recommended length). Every plan, Free included, gets at least one clone. A clone needs a clean recording, so record in a quiet room and avoid background music.
Fix terms before you generate, not after
Technical vocabulary is where narration usually slips. Before generating, scan each script for product names, acronyms and unusual words, and respell the risky ones in the text itself. There is no pronunciation field to fill in, so the fix lives in the spelling. The technique is covered in detail in fixing a mispronounced name.
Keep one running glossary for the course: the original term, the respelled version, and the slide where you confirmed it. Run it as a find-and-replace on each new script and the module 6 narration will say "Kubernetes" the same way module 1 did.
Generate and export audio per slide
Generate slide 1 alone
Paste the script into the [generator](/generate-speech) with your chosen voice. Listen once, fix anything in the text, and regenerate only if you must. Every regeneration costs points again.
Name files for the deck
Download each clip as an MP3 and name it with the slide number: 01-welcome.mp3, 02-goals.mp3. Sorting by name then matches slide order.
Attach audio to each slide
In PowerPoint use the insert-audio-from-file command on each slide. Articulate and similar authoring tools have an equivalent. One file per slide means you can swap a single clip later without touching the rest.
Play the module through
Run the finished module start to finish before publishing. You'll catch a slide where the audio runs past the animation, or a clip that's noticeably louder than its neighbours.
Per-slide files are worth the extra clicking. A single 6-minute file for the whole module looks tidier, but the first time a policy changes on slide 7 you'll be regenerating all six minutes and re-cutting the timing.
What a 10-slide module costs in points
In the app the charge is ⌈characters ÷ 17⌉ points per generation. Here is a realistic module, with slides of different lengths, because real decks aren't uniform.
| Slide | Characters | Points |
|---|---|---|
| 1 Welcome | 500 | 30 |
| 2 Concept A | 900 | 53 |
| 3 Concept B | 900 | 53 |
| 4 Demo | 765 | 45 |
| 5 Worked example | 765 | 45 |
| 6 Concept C | 900 | 53 |
| 7 Quiz setup | 400 | 24 |
| 8 Recap | 600 | 36 |
| 9 Next steps | 400 | 24 |
| 10 Sign-off | 300 | 18 |
| Total | 6,430 | 381 |
381 points is about 6 minutes 20 seconds of audio. Budget for two passes on anything you can't easily redo and the figure becomes roughly 760. At that rate Lite's 3,000 monthly points cover just under 4 modules with re-takes, while Pro's 12,000 cover about 15. A $5 top-up adds 1,200 points that never expire, which is three single-pass modules for a course you build in bursts.
A cheap first test: The Free plan's 60 points pay for slide 1 (30 points) and slide 10 (18 points) with a few left over. Generate those two and you'll hear the voice at both ends of your module before spending anything.
Keep one narrator across a whole course
Consistency breaks quietly. Someone updates module 5 six months later, picks the voice that happens to be at the top of the list, and the course now has two narrators. Stop that with a one-line style sheet stored next to the scripts: the voice, the model, the speed setting and the glossary.
- Write the voice name down. Not "the male voice", the actual voice from your library.
- Note the model and speed. A changed setting alters the pace even with the same voice.
- Store every script as text. Regenerating one slide is cheap only if the script is still around.
- Update by slide, not by module. Replace the single file and leave the other nine.
The same discipline works for any long series, and keeping one voice across a whole channel covers it from the video side. The cost of breaking it is easy to measure: switching voices after module 3 of 8 means regenerating all the slides you already finished.
Where to go from here
If you work in slides, the PowerPoint voiceover generator page goes deeper on narrating a deck.
For LMS modules and course platforms, the e-learning voiceover generator page covers keeping narration easy to update.
E-learning voiceover generator
Narration for modules and LMS courses, with an eye on how easy it is to update later.
PowerPoint voiceover generator
A voiceover per slide for PowerPoint and Google Slides decks.
Frequently asked questions
How long should the narration for each slide be?
Aim for 30 to 60 seconds. In the app that is roughly 510 to 1,020 characters, or 30 to 60 points. A slide that needs more than a minute usually works better split into two.
How many points does a 10-slide module cost?
A module with about 6,400 characters in total costs around 380 points in the app, which is ⌈characters ÷ 17⌉ per slide added together. Budget roughly double if you expect to regenerate slides.
Can I keep the same narrator for the entire course?
Yes. Pick one voice, note its name and your settings in a style sheet, and use the same voice for every slide. You can also clone your own voice from a 15 to 30 second sample and use that for all modules.
How do I fix a technical term the voice gets wrong?
Respell it in the script, test it inside the real sentence, and add it to a course glossary. There is no pronunciation field, so the fix is in the text.
Do I need a paid plan to try it?
No. The Free plan starts with 60 points, enough to generate two or three short slides and hear the voice before you decide.