Reading a teleprompter without sounding wooden mostly comes down to one thing: the script has to follow you, not the other way round. Most flat, robotic reads happen because the speaker is quietly performing to a scrolling speed they guessed before they started — the fix is a mix of delivery habits, script changes and one setup decision, and all of them are learnable in an afternoon.
If you have ever watched a take back and winced at a voice that did not sound like yours, this is for you. Nothing here needs acting training or expensive kit. It needs a script written the way you actually speak, a rig set up so your eyes stay still, and a prompter that does not make you chase it.
The real cause: you are performing to the scroll
Here is the thing nobody tells you when you first prop a script behind a lens. A conventional teleprompter asks you to guess a scroll speed — a words-per-minute number — before you have said a word. Then you spend the whole take serving that number.
Speak a little too slowly and the text creeps ahead of you, so you speed up to catch it. Pause for effect and the pause costs you your place. Stumble, and the script sails on without you while you decide whether to stop or bluff. Your brain is doing two jobs at once: saying the words, and constantly recalculating where you are against a conveyor belt.
That second job is what sounds wooden. The flatness in your voice is not a lack of talent — it is divided attention. The part of you that would normally add warmth, emphasis and natural rhythm is busy doing arithmetic.
You can hear this in the pacing of a struggling read: every sentence comes out at the same speed, because the scroll only has one speed. Real speech does not work like that. You rush the throwaway bits and lean into the point. A fixed scroll flattens both into the same tempo.
There are two ways out. The first is practice — genuinely, people do learn to ride a fixed scroll, the way newsreaders do, and the techniques below will get you most of the way. The second is to remove the conveyor belt entirely. This is the problem TellyPrompter was built around: it listens as you read, and the highlight follows the word you are actually saying — through pauses, ad-libs, skipped lines and backtracking — so the pace of the take is set by your delivery rather than a number you guessed. Speech recognition runs on the device and the audio is never uploaded. Either route works; the rest of this article helps with both.
Write for the mouth, not the page
A large share of wooden reads are wooden before anyone presses record, because the script was written to be read silently rather than said aloud. Written English and spoken English are different registers, and the gap between them is exactly what audiences hear as "reading".
A few changes make a script performable:
Use contractions. "Do not" and "cannot" are fine on the page and stilted out loud. If you would say "don't" in conversation, write "don't". Read any sentence that feels stiff and ask whether you would ever say it to a friend across a table.
Shorten the sentences. A written sentence can carry three clauses comfortably. A spoken one usually cannot — you run out of breath, or the intonation goes strange as you try to signal a comma with your voice. One idea per sentence. Full stops are free.
Break lines where you would breathe. Format the script in short lines rather than dense paragraphs. Each line becomes a natural phrase, and the line breaks do some of your pausing for you.
Say it before you save it. The single most useful edit is reading the draft aloud once and changing every phrase you trip on. If your mouth resists a sentence in rehearsal, it will resist it harder on camera.
None of this is about dumbing anything down. It is about writing in your own speaking voice, which is the voice your audience clicked for.
Eye line and distance: why glancing reads worse than it feels
The camera is unforgiving about eyes. A glance away that feels tiny to you — down at notes, sideways at a second screen — reads on camera as evasiveness or distraction. This is why the physical setup matters more than most delivery advice.
The distances involved are smaller than you might think. If your script sits right next to the lens rather than over it, the angle between "looking at the script" and "looking at the viewer" shrinks as you move the camera back. At arm's length, the flick of your eyes between the two is visible. At two or three metres, it mostly is not.
So: put the script as close to the lens as your setup allows, and put the camera further away than feels natural, then zoom or crop back in. If you use a beam-splitter rig — the mirrored glass that puts the script directly over the lens — the problem disappears entirely, because reading and eye contact become the same act. That is why those rigs exist — and if you are thinking of building one, we have written a full beam-splitter setup guide.
Two practical notes for prompter glass. The mirror flips everything, so make sure your software has a proper mirror mode rather than asking you to read backwards. And the glass eats a surprising amount of light — roughly a third to a half — so favour large text and strong contrast over anything subtle.
Text size matters at ordinary distances too. You want the script readable in your peripheral vision, in glances, without leaning in. Bigger than feels necessary is usually right. Squinting shows.
Rehearse once aloud — then stop
There is a sweet spot between under-prepared and over-prepared, and it is closer to under than most nervous people believe.
Read the script aloud exactly once before you record. Not in your head — aloud, at performance volume, standing or sitting as you will on camera. Two things happen. First, you find the stumbles: the phrases your mouth does not like, the sentence that runs out of breath. Fix those in the script now, rather than discovering them on take four.
Second, you take the edge off the newness. The first read of any script is always the most careful and the most wooden. Spend that read off camera.
Then resist the urge to keep rehearsing. Past one or two reads, something worse than unfamiliarity creeps in: memorisation. Once you half-know the script, you start reciting from memory and using the prompter to check yourself, which produces its own strange cadence — the sound of someone confirming rather than saying. The prompter works best when you trust it completely and let each line be slightly new.
Mark, don't memorise. If a section worries you, put a cue in the script — an underline, a deliberate line break before the hard word — rather than drilling it.
Ad-libbing without losing your place
The moments audiences remember are usually the off-script ones — the aside, the joke that occurred to you mid-sentence, the plainer restatement of a point that came out better than the written version. A teleprompter should permit those, not punish them.
On a fixed scroll, ad-libbing costs you. Every second off script, the text keeps moving, and the price of the aside is a scramble to find your line. Which is why heavily prompted speakers so rarely deviate: not because they have nothing to add, but because the tool taxes it.
If you are riding a fixed scroll, the workaround is to plan your departures: write a short cue like "story about the first attempt" into the script itself, and keep the aside brief enough that the scroll has not travelled far. Some speakers keep a remote or foot pedal to pause the scroll — workable, but one more thing to operate mid-take.
A voice-following prompter changes the economics here: when you wander, the highlight simply waits, and when a phrase from the script comes back out of your mouth, it picks you up again. The aside becomes free.
Either way, give yourself permission in the script. A marked "go off-script here" beat, even just once per video, loosens the whole read — because a script you are allowed to leave stops feeling like a cage.
When to stop polishing and take the take
Perfectionism is the last trap, and it is a delivery problem disguised as a quality standard.
The tenth take of a script is rarely the best one. Fatigue flattens your voice, the words go stale in your mouth, and self-consciousness compounds — by take eight you are no longer performing the script, you are performing "not messing up", which is a different and worse energy.
Some rules of thumb that help:
A stumble is not a failed take. Real speech contains restarts and corrections; leaving small ones in usually reads as human rather than sloppy. If a flub genuinely needs fixing, note where it happened, finish the take, and re-record that sentence rather than the whole script.
Two clean-enough takes beat six perfect attempts. Record the full piece twice and choose in the edit. Deciding "this is a selection problem, not a performance problem" takes enormous pressure off each individual run.
Watch for the fatigue signs. When you start making new mistakes in sections you previously nailed, you are done for now. Tea, walk, come back.
The wooden sound you are trying to avoid is, at bottom, the sound of self-monitoring — a speaker split between saying the words and supervising themselves saying the words. Everything in this article serves the same end: script, setup and software arranged so that supervision is unnecessary, and the whole of your attention is free to just talk to the person on the other side of the lens. That is all sounding natural has ever been.