The text on a teleprompter is rarely written for one. Usually it’s the script the client approved, pasted from Word with the scene headings, the lower thirds and a “(smiling)” still in it.
The presenter sees it for the first time on set, trips over “Q3” in the second line, and the stumble gets put down to nerves.
Reading technique, screen setup and who runs the scroll are in our guide on how to read a teleprompter naturally. This post is about the text you hand the operator the day before.
In short
- Make a separate prompter copy of the approved script and delete everything nobody says.
- Write one thought per line, leave a blank line for each breath and let the screen handle the width.
- Write numbers and abbreviations the way they’re said, and add a respelling in square brackets for hard names.
- Use a few typed marks for pauses and stress, and only where the read-through went wrong.
- Split the text into one block per take, each opening on a line that works after a cut.
Make a prompter version of the script
The script and the prompter text do different jobs. The script is what the client approved and what the edit gets checked against, graphics and all. The prompter text is for one person and holds only what comes out of their mouth.
Make it a separate copy and leave the approved script alone.
From the copy, delete everything nobody says:
- scene headings and shot numbers
- lower thirds, titles and card text
- camera directions and notes like “(see map)”
- music and sound cues
- the speaker’s name in front of each line
- timecodes, durations and margin comments
A nervous presenter will sooner or later read out whatever is left on the screen.
With a two-column script, start from the audio column, as in our video script template, and drop the lines that belong to someone else. The guide to writing a corporate video script covers who says what.
One thought per line, a blank line for each breath
Short sentences are easier to say and easier to pick up after a slip. If the presenter can’t get through one in a relaxed breath, split it.
Then break the lines by meaning. Each line holds one thought, a short sentence or a phrase that belongs together, and a blank line marks where the presenter should breathe or let a point land. With that gap on the screen, nobody needs to type “(pause)”.
Don’t break lines to a fixed width, though. The prompter wraps text to fit its own screen at whatever size the presenter needs, so a line you cut at 40 characters on a laptop wraps again on a tablet and leaves one word stranded below it.
Break where the meaning breaks, let the screen handle the width, and keep names, figures and phrases like “two and a half million” on one line.
Numbers, abbreviations and names, written the way they’re said
A presenter who sees “$2.5M” has to convert it while reading, with the text moving. Some will say “two point five M”. Write what you want to hear:
| In the script | On the prompter |
|---|---|
| $2.5M | two and a half million dollars |
| up 38% YoY | up thirty-eight percent on last year |
| Q3 | the third quarter |
| approx. 1,200 | about twelve hundred |
| 03/04 | March fourth |
| 9:30 a.m. | nine thirty in the morning |
| acme.com/jobs | acme dot com slash jobs |
| e.g. | for example |
Check every date. 03/04 is March 4 in Chicago and April 3 in London, so ask the client before you write the words.
The figures stay as figures in the lower third and the graphics. Spelled-out numbers also keep the word count honest when you time the script against the words that fit in a minute.
Names that could be read two ways get a respelling in square brackets, with the stressed syllable in capitals: “our new office in Wrocław [VROTS-wahf]”. The BBC’s pronunciation guides use the same convention (“pronunciation” is pruh-nun-si-AY-shuhn).
Ask the person how they say their own name instead of guessing, and repeat the bracket in every take where the name comes up.
Before the first take, agree with the presenter that anything in square brackets is never read aloud, and use brackets only for things they need to see and not say.
Marks for pauses and emphasis, and where they stop helping
A few marks are worth using. A slash inside a line gives a short breath (“This year / for the first time”), a blank line gives a longer one, and a word in capitals shows where the stress goes when the meaning depends on it, as in “the FIRST plant in the region to do it”.
They stop helping fast. Put capitals on three words in a paragraph, add four slashes, and the presenter starts performing the marks, with every stress at the same weight.
Mark only what went wrong in the read-through, like a stress on the wrong word or two sentences run together, and leave the rest to punctuation and line breaks.
Bold, italics and the client’s yellow highlight may not survive the paste into the prompter, so stick to marks you can type.
Directions like “(smiling)” come out of the text. If the presenter needs that note, say it before the take.
Split the text into one block per take
Decide before the shoot where the presenter will stop and start again, and split the text there.
Two minutes is about 280 words at a steady pace, which is a lot for someone who reads a prompter once a year. Three or four takes of 30 to 40 seconds are easier on them and give the editor clean ins and outs.
Open each block with a sentence that works as the first thing the viewer hears after a cut. A take that starts with “So, as I mentioned” only works if the previous take makes the edit too.
Mark the top of each block with a bracketed line like [TAKE 2], with a couple of blank lines above it, so the operator finds it fast when the director asks for the top of two.
For someone who freezes even on short blocks, a few points and one exact line can work better than a full text. Our post on employees on camera covers when to drop the prompter.
A before and after
Here’s a paragraph from a 90-second company update, as it came from the client:
SCENE 3 - INT. OFFICE - DAY
LOWER THIRD: Dana Whitfield,
COO
DANA (smiling): Q3 revenue
reached approx. $2.5M, up 38%
YoY, and on 03/04 we open our
new Wrocław office (CUT TO
MAP).
And the prompter version:
[TAKE 1]
In the third quarter,
revenue reached about
two and a half million dollars.
That's thirty-eight percent
more than a year ago.
[TAKE 2]
On March fourth,
we open a new office
in Wrocław [VROTS-wahf].
The scene heading, the lower third, the speaker’s name, “(smiling)” and the cut to the map stay in the script for the edit. “Q3”, “approx. $2.5M” and “38% YoY” are now words, and the date became March fourth after the client confirmed it.
Wrocław has its respelling, and take 2 opens on a line that works straight after a cut.
How this works in RecReview
In RecReview the teleprompter opens from the shoot plan, and each video in the plan is a block with its own text. A video block starts empty.
“Use the task’s script” brings in the task’s latest script, and AI keeps only what’s spoken, dropping on-screen text and lower thirds, camera directions, scene headings, sound and music cues and speaker names. Or choose “Type text” and paste the prompter version you prepared.
Check the result before the shoot day. The cleanup is told to keep only what will be said, so a bracketed respelling or a take marker can go out with the directions.
If it cut too much, the editor can reload the original script in full and you trim it by hand.
The presenter’s screen shows plain text, with line breaks and blank lines exactly where you typed them, and long lines wrap to the screen’s width.
Bold is on or off for the whole text, so emphasis on one word has to be typed, like capitals. The editor counts words and estimates seconds at 140 words a minute as you type.
Applying an edit sends the text to every connected screen, saves it for screens that join later and takes the scroll back to the top of the block, so edit between takes.
Text size, side margin, alignment and bold apply to every screen, in white on black. Screens join by QR code or link without logging in, and each one has its own mirror setting. The teleprompter page has the rest.
