How to stop sounding like you're reading.

Here is the thing nobody tells you when you set up your first teleprompter: the moment you turn on a fixed-speed scroll, you stop being the one who decides how fast this sentence goes. A machine decides. You follow. And following is audible.

The tell is in the gaps, not the words

Listen to someone talk without a script and you hear an irregular rhythm. They hold a beat before the word that matters. They speed through a clause they consider obvious. They breathe in the middle of a thought because that is where the thought turned. The gaps are uneven, and the unevenness is what makes it sound like a person thinking out loud.

Now listen to the same person reading at 90 words per minute of scroll. The gaps flatten out. Every clause gets roughly the same weight. The breaths land at commas rather than at meaning. Nothing is technically wrong, and yet a listener knows within about five seconds that this is being read to them, and quietly downgrades their attention.

That is the actual cost. Not that you sound bad, but that you sound like a recording of a document.

Two failure modes, both from the same cause

A fixed scroll fails in exactly two directions, and you can hear which one you are in.

You are chasing. The text is moving faster than you would naturally speak this particular sentence, so you compress. Your phrasing goes flat because you have no spare time to shape it. You take shallow breaths in the wrong places. Your eyes start tracking ahead in visible little jumps, which the camera catches even at a distance. And you develop a low-grade panic that has nothing to do with the content and everything to do with keeping up.

You are waiting. The text is slower than you want to go, usually because you already know this section by heart. So you fill. You stretch vowels. You over-articulate. You add "so", "right", "and basically". Or you get ahead of the line, realise you have overtaken the prompter, and stall mid-sentence — which is the single most obvious tell on camera.

Most people cycle between the two several times inside one take, because their natural pace is not a constant.

Why "just set the right speed" does not work

This is the advice everyone gives, and it is wrong for a structural reason: there is no single right speed for a script.

Your speed changes constantly, and predictably:

• You slow down on numbers, product names, and anything you might get wrong.
• You speed up on transitions and on material you have said a hundred times.
• You pause after you make a point, because that is what makes it land.
• You slow down again on the first sentence of a new section, while you re-orient.
• You get faster overall as the take goes on and you warm up.

Set the scroll to your average and you are wrong in both directions for most of the script. Set it slow enough to never be chased and you spend the whole video waiting. There is no number that fixes this, which is why fiddling with the speed slider between takes feels like it should work and never quite does.

What helps, whatever tool you use

Some of this is not a software problem at all, and pretending otherwise would be dishonest. Four things that improve delivery independently of any prompter:

Write for the mouth, not for the page. Read your script out loud once before you record. Every place you stumble is a place the sentence is built wrong for speech — usually a subordinate clause, a noun stack, or a sentence with two ideas in it. Cut it in half. Written-English rhythm and spoken-English rhythm are genuinely different, and a script that reads well silently can be almost unspeakable.

Break the text into short lines. One idea per line, generous spacing. This is not decoration. Long unbroken paragraphs force your eye to scan horizontally, and horizontal eye movement is far more visible on camera than vertical. Short lines keep your gaze in a narrow column near the lens.

Mark your breath points before you record. Put a visible break where you intend to pause. If you do not decide in advance, the scroll decides for you, and it will choose the commas.

Get the text as close to the lens as possible. Most of the "shifty eyes" problem is geometry, not performance. A script on a second monitor off to the side will always read as looking away. The further you sit from the camera, the less this matters, which is why the fix is often just backing up and using a longer lens rather than buying anything.

And the part that is a tooling problem

Everything above makes reading better. None of it changes who owns the pace. As long as the scroll runs on a timer, you are still the one adapting.

The inversion is simple to state: the scroll should follow the words you have actually said. Speech recognition tracks your position in the script and moves the text to match. Pause for three seconds and the text waits. Rush a sentence and it keeps up. Improvise a clause that is not in the script and it holds position until you rejoin. You never think about speed, because there is no speed to think about — there is only where you are.

That is the entire reason I built Mirrento, and it is the thing I would test first if you are curious: record the same 60 seconds twice, once at a fixed speed and once with the scroll following you, then listen to the two without watching. The 7-day trial is more than enough time to hear whether your own delivery changes, and it is the only honest way to find out, because this is not a difference you can evaluate from a feature list.

One caveat so this does not read as a sales pitch: fixing the pace does not fix a script that has no point. If you sound bored, the problem is the writing. If you sound nervous, the problem is that you have not said the words out loud enough times. Pace is one specific failure mode of talking-head recording, and it is worth naming precisely so you stop trying to solve it with the wrong tool.