Skip to content

Voice-activated teleprompter: how voice scrolling works and how to get the best from it

A voice-activated teleprompter listens as you read and moves the script to match, so you can pause, slow down or ad-lib without the text running away. Here is how it works, what to look for in an app, and how to write and read so it never loses you.

By our team. Published . 3 sources.

Try it in Cuefold, free

A voice-activated teleprompter listens to you as you read and scrolls the script to keep the line you are saying in front of you. Pause and it waits; speed up and it keeps up; skip a phrase and it catches up when it hears where you are. That makes it easier to sound natural than a prompter scrolling at a fixed speed, which you either chase or wait for. It works best with a script written the way you speak, a quiet room and a microphone close to you, and it is worth knowing what trips it up before you rely on it for a long take.

Fixed-speed scrolling vs voice scrolling

Every teleprompter has to decide when to move the text. There are three ways to do it.

Fixed speedOperator or remoteVoice
Who controls the paceA speed you set in advanceA person, or you with a remoteYour voice
Pausing for effectText keeps moving unless you stop itOperator stops itWaits for you
Changing pace mid-scriptYou have to match the settingOperator follows youFollows you
NeedsNothing extraA second person or a remoteA microphone and a script it can match
Main riskRushing to keep up, or waiting for textOperator attentionLosing your place on ad-libs or noise

Fixed speed works when your pace is steady and you have rehearsed it. Most people read prepared speech somewhere around 150 to 180 words a minute [1][2], but nobody holds one pace for a whole script: you slow down for the key point and speed through a list. That is why a fixed speed so often sounds like reading. Voice scrolling removes the speed setting altogether.

How voice scrolling works

Inside the app, three things happen many times a second:

  1. Speech recognition turns what the microphone hears into words.
  2. Matching compares those words with the script near your current position, allowing for small differences: a recogniser that hears "colour" as "color", a dropped "the", a number said as a word.
  3. Scrolling moves the script so the matched line sits at the reading position, usually the line just under the camera.

The better apps add rules on top. They only move forward when several words in a row match, so one misheard word does not throw the text down the page. They search a little backwards as well as forwards, so if you go back to repeat a sentence, they can follow you back to it.

Where the recognition happens matters too. Some apps send your audio to a server to be transcribed; others recognise speech on the phone itself. On-device recognition keeps working without a connection, adds no network delay and means your words never leave the device. If you record client material, unannounced products or anything private, check which kind you are using before you start.

When voice scrolling is the better choice

  • Talking-head videos you read alone. No operator, and your hands are free.
  • Presentations and speeches you practise. You can stop to think, and the script waits.
  • Long scripts. Small differences in pace add up over five minutes; voice scrolling never drifts.
  • Anyone who changes pace on purpose. If you slow down for emphasis, a fixed speed works against you.

Fixed speed can still be better for music-led pieces, very fast reads, or a room so noisy that recognition struggles.

What trips it up, and how to avoid it

ProblemWhy it happensFix
Text stops on numbers"2026" might be heard as "twenty twenty-six"Write numbers, units and symbols as you say them, unless your app handles them
Text stops on names and jargonThe recogniser does not know the wordKeep going: the next few ordinary words get it moving again
Text jumps aheadA phrase appears twice in the scriptVary the wording of repeated lines
Text waits after an ad-libYour extra words are not in the scriptReturn to the script's words and it catches up
Text freezesBackground noise or music drowns your voiceUse a clip-on microphone and record in a quiet room
Text goes the wrong way after a restartThe app only moves forwardChoose an app that follows a restart back to the line, or scroll back by hand

Write the script for voice scrolling

A script that is easy to match is also a script that is easy to say:

  • One sentence or thought per line. The reading line moves in clean steps instead of mid-sentence.
  • Numbers as words when the app does not handle them: "twenty per cent", not "20%".
  • No stage directions in the text, or put them on their own line in brackets and do not read them. The app is listening for the words you say.
  • Avoid long identical phrases. If two paragraphs open the same way, change one of them.
  • Keep the language consistent. If the app listens in English, a sentence in another language will not match. Some apps let you choose the listening language per script.

Set it up

  1. Put the text under the lens. Viewers are much less sensitive to eyes looking slightly below the camera than to eyes looking sideways [3], so a column right under the lens hides the reading best.
  2. Narrow the column and enlarge the text. Three or four lines visible is plenty; your eyes move less.
  3. Use a microphone near your mouth. A clip-on mic helps the recording and the recognition alike.
  4. Start with a countdown, so the first words you say are script words.
  5. Do a test paragraph and deliberately pause, speed up and restart a sentence. Watch how the text behaves before the real take.

What to look for in an app

  • Recognition on the device, if privacy or working offline matters to you
  • A way to pause and scrub by hand, for when you want control back
  • Text size, column width and font you can change while recording
  • Mirror mode, if you will use it in a glass rig
  • Whether it records video itself, or only shows the script
  • What it does when you repeat a line: stay put, jump, or ease back to it

That last point matters more than it sounds. When you flub and restart a sentence, a prompter that only moves forward leaves the line you are repeating above the reading position. One that follows you back keeps the line in front of you, so the second read is as easy as the first.

Quick checklist

  • One sentence per line, numbers written as said, no repeated phrases
  • Text under the lens in a narrow column
  • Microphone close, room quiet
  • Test paragraph with a pause, a speed change and a restart
  • Know how your app handles restarts before the real take

Common questions

What is a voice-activated teleprompter?
A teleprompter that listens as you read and scrolls the script to match your voice, instead of moving at a fixed speed. It waits when you pause and keeps up when you speed up.
Does a voice teleprompter work offline?
It depends on the app. Apps that recognise speech on the device work without a connection, and your audio never leaves the phone. Apps that send audio to a server need a connection.
Why does my voice teleprompter keep stopping?
Usually because it cannot match what it hears to the script: numbers written as digits, unusual names, background noise or words you added. Write numbers as you say them, use a microphone close to you and return to the script's words after an ad-lib.
Is voice scrolling better than a fixed speed?
For most solo talking-head videos and practice runs, yes, because you can change pace and pause without the text running away. Most people read at around 150 to 180 words a minute [1][2] but vary within a script, which a fixed speed cannot follow.
What happens if I restart a sentence?
Some apps stay where they are, which leaves the line you are repeating above the reading position. Better ones follow you back to the start of the line. Test this before you record.

Sources

Checked on 11 November 2026.

  1. 1.ACX Help Center, "How long will my narrated audiobook be?" (about 9,300 words an hour): https://help.acx.com/s/article/how-long-will-my-narrated-audiobook-be
  2. 2.Rodero, E. (2012), "A comparative analysis of speech rate and perception in radio bulletins", Text & Talk 32(3), 391-411: https://doi.org/10.1515/text-2012-0019
  3. 3.Chen, M. (2002), "Leveraging the asymmetric sensitivity of eye contact for videoconferencing", CHI 2002, Stanford University: https://graphics.stanford.edu/papers/eye_contact/