You have probably watched someone you like record a video and seen the moment it goes wrong. Ten minutes earlier, over coffee, they were funny and quick. Then the camera comes on and they say, "In today's presentation, I would like to discuss several considerations regarding our approach." Nobody has ever said that sentence to a friend. The person you liked has left the room, and a stranger in a blazer is standing where they were.

Most people blame delivery. They think they need more energy, a better voice, or more confidence. But the damage usually happens earlier, at the keyboard. They wrote an essay and then tried to perform it.

The page and the ear are different machines

A reader controls the pace. They can slow down on a hard sentence, jump back, skim ahead, or stop to think. A listener can do none of that. Speech moves past once and is gone, and the listener has to build meaning in real time.

The linguist Wallace Chafe studied how people actually talk and found that spontaneous speech comes in short bursts, which he called intonation units. Each one is a few words long and carries roughly one new piece of information. Written prose does the opposite. It packs clauses inside clauses, stacks modifiers, and holds a subject in suspense for twenty words before the verb arrives. That is fine on a page. In a voice, it forces the listener to hold the whole structure in memory while they wait for it to resolve.

This is why a well-written paragraph can sound awful aloud. The writing is good for a reader and the format is wrong for a listener.

What spoken language does that written language doesn't

Look at a transcript of your own casual speech and a few habits stand out. They are the same habits a script has to copy.

Short units. You rarely say more than one idea before a breath. "I almost quit. Then something small changed." That is two beats. The written version would be "After nearly quitting, I experienced a minor shift that altered my trajectory."

Repetition. On a page, repeating yourself looks lazy. In speech it is how people keep their place. Restating the key point in different words gives a listener a second chance to catch it.

Contractions and fragments. "It's not that hard." "Not even close." "Which is the point." Fragments are how people sound when they are thinking in front of you.

Direct address. Spoken language leans on "you." It is a conversation with one person, even when ten thousand people watch.

Concrete nouns. You say "the red notebook" and not "organizational materials." Abstractions take effort to decode, and a listener has no spare effort.

The curse of knowledge makes it worse

There is a second reason scripts drift into stiffness, and it is more subtle. In a well-known 1990 Stanford study by Elizabeth Newton, people were asked to tap out the rhythm of a familiar song on a table while a listener tried to guess it. The tappers predicted the listener would get it right about half the time. The listeners almost never did. The tappers could not stop hearing the melody in their heads, so they could not imagine its absence.

When you write about something you know well, you are the tapper. You skip steps, use shorthand, and assume context. On the page that costs you some clarity. In a video it costs you more, because the viewer can't pause you to ask what you meant. A script that works has to be written for someone who has never heard the tune.

A practical way to write: talk first, then tighten

The most reliable method is also the least glamorous. Don't open a blank document and try to be eloquent. Say it first.

Explain your topic out loud to an imagined friend, or to your phone's voice memo, and don't worry about polish. Transcribe it. That rough, rambling version contains your real rhythm, your real examples, and the phrases you naturally reach for. Then edit it down, cutting the detours and keeping the sentences that sound like you.

The usual instinct is to write something impressive and then make it sound casual. That runs the wrong way. It is much easier to make talk tidy than to make prose human.

The read-aloud test

Once you have a draft, read it aloud at normal speed, and listen for the places where your body objects.

  • If you run out of breath mid-sentence, the sentence is too long. Split it.
  • If you stumble on the same phrase twice, it is too written. Replace it with what you would actually say.
  • If you catch yourself changing a word as you read, take the change. Your mouth just edited your script for free.
  • If you feel a flicker of embarrassment, the line is probably performing instead of communicating. Cut it or soften it.

Conversational speech usually lands somewhere around 130 to 160 words per minute, though it varies a lot by person and topic. Timing your read gives you a rough length, and pacing yourself against it beats guessing.

Build the script around beats, not paragraphs

Format matters here too. A script isn't a document, and it shouldn't look like one. Break it into one-breath lines, so the shape of the speech is visible before you say a word. Put each new idea on its own line. Leave white space where you want a pause. Mark the one word in a sentence that deserves weight.

This also protects you from a common failure. In a dense block of text, your eyes lose their place and your delivery goes flat. In short lines, you can glance down, take in a thought, and deliver it to the lens.

Your next moves

  • Record a two-minute voice memo today explaining your video topic to a friend who knows nothing about it. Transcribe it with your phone's dictation feature. This is your raw material, and it already sounds like you.
  • Rewrite your opening line three ways, then pick the one you would actually say to someone standing in your kitchen. If it begins with "Welcome to" or "In this video I will," try again.
  • Run the read-aloud test on your current draft. Mark every spot where you stumble, run out of breath, or change a word. Fix each one before you touch anything else.
  • Break your script into one-breath lines. Put a single idea on each line, and leave a blank line wherever a pause belongs.
  • Swap one abstraction for a picture. Find the vaguest sentence in the script ("improve your workflow") and replace it with a specific scene ("the Sunday night you spend dragging files into folders").

Where a teleprompter fits

A script written for the ear is easier to read well, but reading it is still its own skill. That is where we built Cue. It is a teleprompter that follows your voice, so the text keeps pace with you instead of the other way around, and your eyes stay on the lens. When the words already sound like you, that is a good place to start. If you're curious, you can see how it works at cue.lumenlabs.works.