Voice activated teleprompter: how it works, and four things to test before you trust one

A fixed-speed teleprompter makes you keep pace with it. You set a speed before the take, then spend the take managing it: rushing to catch up, or waiting for the words. A voice activated teleprompter turns that around. It listens as you read, works out where you are in the script, and moves the words to meet you. Stop to breathe and it waits. How well that works varies between apps, and it shows in the moments that matter: a stumble, an ad-lib, a second try at a line. Below: how it works, where your audio goes, and four tests to run before a real shoot.

How a voice activated teleprompter works

Under the hood there are two jobs.

First, speech recognition. The app listens through your microphone and turns what you say into text, continuously, as you talk.

Second, matching. It compares that running text against the script to work out where you are. This is the harder part and the one that separates good apps from bad ones. People don’t read scripts word for word. They swap a word, drop a “the”, run two sentences together, stumble and restart. A matcher that needs an exact match stalls every time you do any of that.

Once it knows your position, the app scrolls so the line you’re reading sits at a fixed point on the screen, ideally right next to your camera lens so your eyes stay where the viewer is.

Where your audio goes

This is the question worth asking first, and most people don’t.

Speech recognition can run in two places:

• In the cloud. Your microphone audio is streamed to a server, which sends back text. This is how some browsers’ built-in speech recognition works, and it means your script, and anything you say near the microphone, is being sent somewhere.
• On your device. A speech model is downloaded once and runs locally. Nothing you say leaves your computer.

For a presenter rehearsing a product launch, an internal announcement, or anything under embargo, that difference matters. It’s worth checking which kind your teleprompter uses, and not just taking the marketing page’s word for it.

In Miracue, the speech model runs entirely on your device. It downloads once, then starts instantly after that, even offline. You don’t have to trust that claim: open your browser’s developer tools, watch the network panel while you read, and you’ll see no audio going anywhere.

Four tests to run before you trust one

Most voice teleprompters handle the easy case well: you read the script straight through at a steady pace, and it scrolls along with you. That tells you very little. Real takes aren’t like that. These four tests are.

1. Change your pace mid-sentence

Read a sentence slowly, then speed up through the next one. Then slow right down again.

What good looks like: the script tracks you both ways without lurching. It shouldn’t overshoot when you speed up, or keep drifting after you slow down.

2. Stop completely

Read a line, then stop and count to five.

What good looks like: the script holds still. It doesn’t creep forward, and it picks up from the right place when you start again. This is the test that matters for breathing, for a nervous pause, and for a director shouting “hold on” mid-take.

3. Go off script

Read the first line, then say something that isn’t in the script at all. Talk for fifteen or twenty seconds, then rejoin the script.

What good looks like: while you’re off the page, the script doesn’t scroll. When you come back to the words, it finds you. A weak matcher either keeps scrolling as if you were still reading, or loses you completely and has to be restarted by hand.

This matters more than it sounds. The best takes often come from a presenter who found a better way to say something in the moment. A teleprompter that punishes that trains people to stop doing it.

4. Fluff a line and retake it

Read a paragraph, fumble a line, then take the sentence again.

What good looks like: one press gets you back to the start of the line, without stopping the take.

A retake is the moment most prompters end the take: you reach for the keyboard, scroll back by hand, and lose the energy you had. Look for a back button on a remote, and chapters you can jump to mid-read.

In Miracue, the clicker’s previous button (or Page Up) jumps back three lines, a chapter button jumps back a section, and voice follow carries on from wherever you land.

What else to look for

It shouldn’t need exact words. A changed phrase or a skipped line shouldn’t stall it.

You should be able to tell it’s listening. A clear indicator that the microphone is live, ideally one that moves with your voice, tells you at a glance whether it’s hearing you or merely switched on.

A way to mute it without touching the screen. On a shoot you want a key or a clicker button that pauses listening, so you can talk to the crew between takes without the script racing off.

Honest limits. Voice following leans on your device. It works best on a computer or a recent iPad, while scrolling and remotes work on every device. A trustworthy app tells you where it runs best rather than letting you find out mid-shoot.

Language. Check that it supports the language you present in. Miracue’s voice following is English only today.

How the options compare

Several teleprompters offer voice scrolling. From their own sites, checked in September 2026:

• PromptSmart calls its feature VoiceTrack, which it says “automatically scrolls as you speak, stops when you pause or improvise, and seamlessly resumes when you return to your script.” It runs on iOS, Android, the web and PC, and is paid, with a seven-day trial.
• Teleprompter.com calls its feature VoiceGlide: the text “follows your voice, pauses when you pause, and picks up where you left off.” There’s a free tier, with a watermark on recordings.
• Speakflow lets you “scroll using your voice or manually.”
• Miracue follows your voice on a free plan with no account needed, runs the speech model on your device, and waits when you stop. It’s browser-based, so it runs on a computer, tablet or phone without an install.

The claims on those pages sound similar, which is exactly why the four tests above are worth ten minutes. Put the same paragraph through each app you’re considering and see which one keeps you in the take.

Setting one up for a shoot

1. Put the reading line next to the lens. Move the line you’re reading to wherever your camera actually is: high on the screen if the camera sits above it, centered if you’re using a beam splitter. The whole point of a teleprompter is that the viewer can’t tell you’re reading.
2. Use a decent microphone. The one on your laptop works, but a headset or lavalier gives the recognition cleaner audio, especially in an echoey room.
3. Rehearse once with voice following on. It lets you find the lines where you naturally speed up or pause, before the camera is rolling.
4. Keep a manual fallback. A clicker or keyboard shortcut to pause and jump sections means a problem never ends the take.

If you want to try it, Miracue runs in your browser. Paste a script, turn on voice follow, and run the four tests on it.

Frequently asked questions