Voice-Follow vs Auto-Scroll: Which Teleprompter Mode to Use When

Most teleprompter advice treats scrolling as a settings question — find the right speed, save it, done. That works right up until the first time you pause to think, and the script sails calmly on without you.

There are really two different philosophies here, and they fail in opposite directions. Auto-scroll moves the text at a constant speed and expects you to keep up. Voice-follow listens to what you are saying and moves the text to match. Neither is better in general. They are better at different jobs, and once you can name which job you are doing, choosing takes about three seconds.

Here is the honest comparison, including the cases where each one is the wrong choice.

What each mode actually does

Auto-scroll is the older idea and the simpler one. You set a speed with a slider, tap record, and the text creeps upward at that rate forever. It is completely predictable. It is also completely indifferent to you.

Voice-follow inverts the relationship. In Teleprompter: Camera Overlay, the app listens through the microphone, matches what you are saying against the script, and advances the text to wherever you actually are. Speed up and it speeds up. Pause mid-sentence for four seconds and it waits. Lose your place and stumble back a few words, and it finds you again rather than carrying on without you.

Worth knowing about the listening part specifically: it uses Apple's on-device speech recognition. Nothing is uploaded, the whole thing works in airplane mode, and it auto-detects the script's language — every language iOS supports for on-device recognition works, not just English. That last detail matters more than it sounds, and I will come back to it.

When voice-follow is clearly right

Voice-follow earns its keep whenever your delivery is not going to be metronomic — which is most of the time, if you are trying to sound like a person.

• You want natural pacing. Real speech slows down for the important sentence and speeds through the setup. Auto-scroll punishes both.
• You improvise off the script. If you habitually add a half-sentence that is not written down, voice-follow tolerates it and picks you up again.
• Long takes. Over four or five minutes, tiny speed mismatches compound. By the end of an auto-scrolled long take you are either racing or waiting.
• You are nervous. Nerves change your speed unpredictably, usually upward. Being chased by text makes that worse in a feedback loop.
• Unfamiliar material. Reading something you wrote an hour ago is slower and more halting than reading something you know.

The second language case deserves its own mention. Recording in a language you speak well but not natively means your pace varies far more than it does in your first language — some phrases come out fluently, others need a beat. A fixed speed set for your good sentences will steamroll your careful ones. This is probably the single strongest argument for voice-follow, and it is why the language auto-detection matters rather than being a spec-sheet line.

When auto-scroll is the better tool

This is the part most app write-ups skip, because auto-scroll is the boring old feature. But there are real situations where a constant speed is the correct answer, and pretending otherwise helps nobody.

• You need a specific duration. If the video has to be sixty seconds, a known scroll speed is a timing instrument. Voice-follow will happily let you run to seventy.
• Noisy environments. Speech recognition needs to hear you. A cafe, a busy street, or a room with a loud fan degrades the matching.
• Silent or lip-synced recording. If you are not speaking aloud at all — recording B-roll to voice over later, or filming mouthed sections — there is nothing for voice-follow to follow.
• Very short, heavily rehearsed scripts. For a twenty-second piece you have already run four times, the constant speed is one less thing behaving unpredictably.
• Deliberate pace training. If you know you rush, a fixed scroll speed set slightly slow is a genuinely useful discipline. Voice-follow will accommodate your rushing instead of correcting it.

That last one is the honest trade-off with voice-follow generally: by adapting to you, it removes a constraint that was sometimes doing useful work. If your problem is that you talk too fast, a mode that speeds up when you do is not going to fix it.

The failure modes, side by side

Both modes break. What separates them is how they break, and how recoverable it is mid-take.

Auto-scroll fails gradually and unrecoverably. You fall half a line behind, then a full line, and the only fix is stopping. You cannot speed yourself up to catch a scroll without it being audible — that panicked acceleration is one of the most recognisable sounds in amateur video.

Voice-follow fails suddenly and recoverably. If it mishears a phrase it may hesitate or jump; but because it is matching against your actual words, saying the next few words clearly usually snaps it back. The recovery is a second of awkwardness rather than a ruined take, and it is often invisible after a cut.

In practice I would rather have occasional one-second glitches than a slow inevitable drift, which is why voice-follow is the sensible default. But note the word default. It is not the answer to everything.

A quick way to decide

Two questions, in this order:

• Is the room quiet enough that a voice assistant would understand you? If no, use auto-scroll. Nothing else in this decision matters.
• Does the video have a hard time limit? If yes, lean auto-scroll and rehearse to the clock. If no, use voice-follow.

That covers the vast majority of real recordings. Everything else is preference.

Settings that matter more than the mode

One thing worth saying plainly: people spend a lot of time agonising over scroll behaviour and almost none on the two settings that visibly change the footage.

The first is text size. Larger text means fewer words on screen, which means your eyes travel less, which means less visible scanning on camera. Most people set it too small because they want to see more of the script, and then wonder why their eyes look busy.

The second is where the text sits. The whole point of a camera-overlay teleprompter is that the script floats over the preview right next to the front lens, so reading and looking at the lens are nearly the same action. If the text is drifting to the bottom of the screen, no scrolling mode will save the eye contact.

And regardless of mode, the script never appears in the saved video — it is an overlay on the preview only, so you can record in portrait 4K with a wall of text in front of you and the finished file is clean.

Try switching mid-project, not mid-take

A practical habit: pick the mode per video rather than setting it once and forgetting. The same person filming a timed sixty-second ad and a five-minute explainer on the same afternoon genuinely wants different behaviour for each.

If you are new to voice-follow, the cheapest way to build trust is to record the same ninety-second script twice, once in each mode, and watch both back. The difference in your eyes is usually more obvious than the difference in the audio.

If neither take sounds right, the problem may not be the scrolling at all — badly built sentences are hard to deliver in any mode, and writing a script that does not sound written fixes more than settings ever will. For the broader case on using a prompter in the first place, there is also the benefits of using a teleprompter app.

Next in this series: teleprompter technique for Instagram Reels specifically — why the vertical crop changes where you should put the text, and what a nine-second hook does to your scripting.