AI Narrator — One Voice, For The Whole Series
Documentary weight, museum calm, storytelling warmth or plain explainer clarity. Pick the register, hand over a script or just the subject, and get narration that sounds identical in episode forty as it did in episode one.
An AI narrator is a text-to-speech voice used to carry spoken narration through a video, course, documentary, story or audio guide. You supply the script — or a description of what it should cover — pick a voice, and get an audio file to lay under your edit. The advantage over recording a person is not cost so much as consistency: the voice does not change between sessions, so material made months or years apart still sounds like a single production.
15 free credits on signup — enough for one complete track with cover art. No card required.
Eight Things That Need Narrating
Narration is one word for very different jobs. What is being narrated decides the register, the pace and how much personality the voice is allowed to have.
Documentary
OnyxAuthoritative and unhurried. Carries factual copy over pictures without competing with them.
Explainer video
AlloyNeutral and even. The narration should be the thing you do not notice while you follow the diagram.
E-learning module
SageCalm and measured, and re-recordable one sentence at a time when the course content changes.
Story & fiction
FableExpressive and warm. Enough colour to hold a chapter without turning into a performance.
Museum & audio guide
SageSlow, clear, and written to be heard standing up in a room full of other people.
Corporate & internal video
EchoSteady and neutral. Deliberately unremarkable, which is correct for a policy or process film.
Children’s & bedtime
ShimmerSoft and gentle, at a pace that assumes the listener is falling asleep on purpose.
True crime & history
OnyxWeight without melodrama. The material supplies the tension; the voice should not add any.
How to Generate Narration in 3 Steps
Register first, script second, and reuse the voice forever after.
Choose the register before the words
Weight, calm, warmth or clarity — decide which the material needs. A museum guide and a true-crime documentary are opposite ends of that list, and the script should be written to match.
Paste the script, or describe the piece
Up to 4,000 characters, about four minutes. If you only have the subject and the audience, describe those instead and the narration gets written for you first.
Generate, then keep the voice for the series
It renders in about a minute as a 24 kHz WAV. Note which voice you used — reusing it means every future episode or module matches without any effort.
Write It To Be Heard, Not Read
Narration that came from a document always sounds like it. Sentences that work on a page run out of breath aloud, and there is no SSML here to rescue them — punctuation is your only pacing control.
“The canal, which was completed in 1804 after a series of funding delays that nearly ended the project entirely, transformed the region’s economy.”
“The canal opened in 1804. It had nearly been abandoned twice. Within a decade it had changed the economy of the entire region.”
“In this module we will be examining the various considerations relevant to workplace risk assessment.”
“This module covers risk assessment. What to look for. Who is responsible. And what to write down.”
“Please note that the exhibit before you dates from the late Bronze Age period.”
“The bowl in front of you is three thousand years old. Look at the rim. Someone repaired it.”
The Case For A Voice That Never Changes
Narration is usually the part of a project that outlives the session it was recorded in. You add a module, correct a fact, extend a series — and with a human narrator each of those means booking them again, matching the room, and hoping they have not had a cold. A generated narrator has no session to rebook. The voice in your twentieth video is bit-for-bit the voice in your first.
- Correct one sentence without re-recording the whole section
- Extend a series months later with narration that still matches
- Uncompressed 24 kHz WAV, so a timeline export is not re-encoding an MP3
- Script written from a brief when all you have is the subject and the audience
- Honest limits: 4,000 characters per pass, no SSML, no voice cloning
Narration Briefs To Start From
Paste your own script instead, if you already have one.
Documentary segment
“A four-minute documentary narration about a village that was flooded to build a reservoir in 1953, factual and unhurried, present tense where possible, no music cues and no dramatic language”
Explainer video
“A neutral three-minute narration explaining how a heat pump moves warmth rather than making it, for homeowners with no technical background, short sentences, one idea per sentence”
Course module
“A calm introduction to a module on giving feedback at work, covering why it goes wrong, three things to do instead, and what to say when it lands badly”
Museum audio guide
“A slow ninety-second guide to a Bronze Age bowl in a display case, describing what to look at, the repair on the rim, and what that tells us about who owned it”
Story reading
“A warm, expressive narration of a short story opening about a lighthouse keeper who receives a letter with no postmark, restrained delivery, no character voices”
Corporate video
“A steady internal narration introducing a new expenses process, covering what changes, when it starts, and where to find the form, plain and free of enthusiasm”
Who Uses An AI Narrator
Video producers
Documentary and explainer narration cut straight into a timeline as an uncompressed WAV.
Course creators
Module narration that can be corrected a sentence at a time when content changes.
Storytellers & authors
Warm, expressive readings for excerpts, short fiction and bedtime content.
Museums & venues
Audio guides recorded once, extended later, matching across every exhibit.
What You Get
24 kHz WAV narration
16-bit mono, uncompressed, up to about four minutes per generation.
The script alongside
Saved with the audio, so a correction is one edit and one regeneration.
Nine registers
Swap voice and regenerate to hear the same script land differently.
Share link
Send narration for approval before anyone else creates an account.
Frequently Asked Questions
What producers and course makers ask before committing a project to one voice.
What is an AI narrator, and how is it different from a voiceover?
In practice the words overlap, but narration usually means carrying the material itself — a documentary, a course, a story — where a voiceover is often a short layer over something else, like an advert or a social clip. The distinction matters for one reason: narration runs long and is heard continuously, so consistency and pacing matter more than personality. That is exactly where a synthetic voice does well, because it never gets tired at minute nine.
Which voice makes the best narrator?
It depends on the material and it is worth being specific. Onyx is deep and authoritative — documentary, history, anything factual that needs weight. Sage is calm and measured for reflective and instructional material. Fable is expressive and warm, which is the storytelling one. Shimmer is soft and gentle for sleep and children’s content. Alloy is neutral and even, the safe default for explainers where the narration should be invisible.
How much can it narrate at once?
Up to 4,000 characters per generation — roughly 650 words, or about four minutes of narration. For most explainer videos, course modules and documentary segments that is a whole section in one pass. Longer pieces get generated in parts, and because the voice does not vary between runs, the parts join without an audible seam.
Do I need a finished script?
No. Paste one if you have it. Otherwise describe the piece — "a four-minute narration explaining how canal locks work, for a museum audio guide, plain and unhurried" — and the script is written before it is spoken. That is the practical difference from a plain text-to-speech box, which cannot help you until the words already exist.
Can I add pauses, emphasis or pacing marks with SSML?
No. SSML is not implemented, and there are no break tags, emphasis marks or speed controls. Punctuation is the only pacing tool you have, which is genuinely enough for narration if you use it — end sentences where you want a breath, and split long ones. Narration written in short sentences reads better out loud anyway.
Will a narration recorded today match one I add next year?
Yes, and that is one of the better reasons to use a synthetic narrator for a series. The voice is identical every time, so a course you extend in eighteen months, or a documentary series you add an episode to, still sounds like one production. Re-recording a single corrected sentence is trivial for the same reason — with a human narrator you would usually re-record the whole section to match.
Is it free, and what does narration cost?
Signup gives you 15 credits and asks for no card. A narration is 10 credits, or 15 with AI cover artwork. Playback and a public share link are free; downloading the WAV needs a paid plan from $15 a month for 250 credits — about 25 narrations, or roughly an hour and a half of finished audio.
Can I use AI narration in a monetised video or a paid course?
Yes, on a paid plan, which carries full commercial usage rights. Because the voice is generated rather than performed, there is no talent to re-clear when a video keeps earning or a course is resold, and no usage term that quietly expires two years in.
More Narration & Voice Tools
Same studio, same credits — a different thing to read aloud.
Pick The Register. Keep It Forever.
Narration that still matches when you add episode forty. Free to start — 15 credits on signup, no card needed.