Using a Teleprompter for Voiceovers and Podcasts: When the Camera Is Off, the Rules Change

Most teleprompter advice is about eye contact: keep the text near the lens, do not let your eyes track left to right, look like you are talking rather than reading. Switch the camera off and almost all of that stops mattering. A voiceover, a podcast cold open, an audiobook sample or a narrated screen recording has exactly one job, which is to sound right, and a teleprompter used for audio has a different set of rules. This post is about those rules.

It is written around Teleprompter: Camera Overlay, which is built for on-camera use, but the ideas apply to any prompter: how to format a script for the ear, which scrolling mode to use when nobody can see your eyes, how to record in takes, and the honest cases where you should put the script down and talk.

What changes when the camera is off

• Eye line is irrelevant. You can hold the phone wherever your voice sounds best, usually below the mouth and a little to the side so breath does not hit the microphone.
• Text size can be huge. With no need to keep the script near a lens, use the largest text the screen allows. Fewer words per line means fewer stumbles.
• Pauses are free. On camera, a pause while you find your place looks like a freeze. In audio, a pause is a breath and an edit point. You can stop, re-read, and start again mid-sentence without a visible seam.
• Pace is the whole performance. Listeners hear rushing immediately. The script should be marked for speed and emphasis in a way you would never bother with on video.

Voice-follow or auto-scroll for audio work?

The voice-follow vs auto-scroll question has a different answer for audio than for video. On camera, voice-follow wins because it lets you pause and the text waits. For voiceovers the answer depends on the kind of read.

Use voice-follow for conversational reads: podcast intros, explainer narration, anything where you want to ad-lib a sentence, laugh, or restart a line. The app listens on-device, follows what you actually say, and finds your place again if you stumble, so a restarted sentence does not leave you scrolling back by hand.

Use auto-scroll for timed reads: a 30-second ad spot, a narration that has to land on specific visuals, or a meditation track that must not speed up. Set the speed slider so the script finishes exactly at the target length, then make yourself match it. The scroll becomes a metronome, and that discipline is the point.

One caveat specific to audio: if you record with the phone's own microphone, voice-follow is using the same microphone to listen, which is fine, but a very quiet whisper read may not register well enough to track. Speak at normal volume or switch to auto-scroll for whispered or ASMR-style work.

Format the script for the ear, not the eye

A voiceover script should look odd on paper. Short lines. One idea per line. Marks for breath and emphasis. Here is a before-and-after for a podcast cold open.

Before: "Welcome back to the show. Today we're talking about why most people quit learning a language after three months, and what the ones who don't have in common, which turns out to be surprisingly simple."

After: "Welcome back. / Today: why most people quit a language after three months. / And what the ones who DON'T have in common. / (beat) / It's simpler than you'd think."

• Use a slash or a line break for every breath. Your eyes read ahead; your lungs do not.
• Capitalise the one word per sentence that carries the meaning. Not three. One.
• Write "(beat)" where you want silence. On a prompter, silence has to be written down or it does not happen.
• Spell numbers and names the way they are said: "twenty twenty-six", "ess-cue-ell". The read will be cleaner and the voice-follow tracking will be too.
• Cut anything you would not say to a friend across a table. The script that doesn't sound written rules apply double when there is no face to carry the words.

A recording setup that takes two minutes

• Record into whatever you normally use for audio. The teleprompter app records portrait video, which is useful if you want a video version too, but for pure audio you can run the prompter on the phone and record on a laptop or a dedicated recorder.
• Put the phone on a stand at mouth height, 30 to 40 centimetres away, slightly off-axis from the microphone. Do not hand-hold it; the small tilts show up as level changes in your voice.
• Turn text size up until there are four or five words per line. Text size matters more here than on camera because you are reading continuously rather than glancing.
• Silence notifications on the phone. A banner mid-read is a retake.
• Record a ten-second test and listen back with headphones before the real take. Room noise and mouth clicks are easier to fix in the setup than in the edit.

Record in takes, not in one pass

The single biggest improvement for most people is to stop trying to read the whole script in one perfect take. Break it into sections of 30 to 60 seconds, record each as its own take, and keep the good ones. Teleprompter: Camera Overlay keeps a takes library, so each section's recording stays attached to the script; if you are recording audio separately, name the files by section so the edit is a matter of assembling rather than searching.

Within a take, if you fluff a line, do not stop. Pause, say the line again from the start of the sentence, and keep going. With voice-follow the text waits for you; with auto-scroll, let it run and pick the sentence up again when the line comes round, or end the take and redo just that section. Your editor (even if that is you, tomorrow) will cut at the pause. A three-second silence is far easier to remove than a restart from the top.

When not to use a teleprompter for audio

There are kinds of audio where reading makes things worse, and it is better to know them than to fight them.

• Interview podcasts. Script the intro and the questions you must ask; do not script your reactions. Listeners can hear a read response from a mile away.
• Anything you know cold. If you have given the talk fifty times, a prompter will make you sound like it is the first. Use bullet points on the screen instead of sentences.
• Highly emotional reads. Narrating something personal from a script tends to flatten it. Record it from memory first, then use the script only for the facts you got wrong.
• When the script is still changing. A prompter rewards a finished script. If you are still thinking, talk it through unscripted, transcribe it, and then write the script from your own words.

A note on privacy for voice work

Voiceover scripts are often confidential: an unreleased product, a client's ad copy, a course that has not launched. Teleprompter: Camera Overlay's voice-follow uses Apple's on-device speech recognition, works in airplane mode, and uploads nothing, which means the script and your voice never leave the phone. If you are reading something under NDA, that is worth checking in whatever prompter you use, because several cloud-based ones send audio to a server to do the same tracking.

The app is free with occasional ads and a one-time purchase to remove them, there is no subscription and no watermark, and it is iOS-only, so the audio-first workflow above is an iPhone workflow. If the rest of your audio chain is on a Mac or a PC, the phone simply becomes the script display, which is honestly the cleanest way to use it.

Next in the series

Audio was the easy case because the camera was off. Next we put it back on and go through the full talking-head filming setup: where the light goes, where the phone goes, how to frame for the prompter overlay, and the five-minute checklist that makes a batch of videos look like they were shot by the same person on the same day.