you're about to film a talking-head video. a YouTube upload, a short, a lesson for a course, a demo of the thing you built. you've never written a script, and the last time you tried winging it you got four minutes of "um" and one usable sentence.
here's the method i use. it goes from a one-line idea to a script you can read on camera without sounding like you're reading. no theory, no three-act structure diagrams. six steps, a worked example, and the length math.
the short version: a video script is written for the ear, not the eye. everything below follows from that.
what a video script actually is#
a blog post gets skimmed. a script gets said, once, at the speed of your mouth, to someone who can't scroll back. that changes the rules.
on a page, a long sentence with two commas and a clause in the middle is fine. the reader's eye handles it. out loud, that same sentence runs you out of breath and loses the listener halfway. on a page, you can say "as mentioned above." in a video, there is no above.
so a good script looks a little strange written down. short sentences. some fragments. the same idea said twice in different words. contractions everywhere. it reads like a transcript of someone talking, because that's what it's meant to become.
how to write a video script, step by step#
1. write the one-line idea#
before any script, write one sentence that says what the viewer gets. not the topic. the takeaway.
"home office audio" is a topic. "a rug and a bookshelf will fix your echo better than a new mic" is a takeaway. if you can't write that sentence, you aren't ready to script. you're ready to think some more.
this line does two jobs. it keeps you from wandering, and it's usually the raw material for your hook.
2. outline in three to five beats#
a beat is one thing you say before moving on. a 60 second video has room for three. a 5 minute video might have five or six. more than that and each one gets too thin to land.
write the beats as bullets, a few words each:
- the problem (your room echoes)
- why (hard surfaces bounce sound)
- the fix (soft stuff, behind the mic and in front of you)
- the test (clap and listen)
that's the skeleton. you'll fill it in next, but you'll never get lost, because you know where every section ends.
3. draft it out loud#
this is the step people skip, and it's the one that matters most. don't type the script and then read it. say it, then type what you said.
talk through each beat as if you're explaining it to a friend across the table. record a voice memo if it helps. then write down the version that came out of your mouth, not the version your inner essayist wants. you'll notice spoken sentences are shorter, more direct, and use smaller words. keep all of that.
if you draft with ChatGPT or Claude instead, tell it to write in short spoken lines, then read the result aloud and rewrite anything that trips you. the llm workflow post covers the prompts that help.
4. put the hook first#
once the body exists, go back and find the sharpest thing in it. move it to line one.
the first line should be the first thing your viewer hears. not "hey everyone." not your name. the claim, the tension, the specific number. on a short you have about three seconds before the thumb moves, and on a longer video the opening still decides whether people stay for the rest. writing hooks that survive the first 3 seconds has patterns you can borrow.
write the hook last. you can't promise a payoff until you know what the payoff is.
5. cut for the ear#
now edit, with your mouth, not your eyes. read the whole thing aloud at the pace you'll film it. anywhere you stumble, the script is wrong, not you.
the usual fixes:
- split long sentences. if you need a breath in the middle, put a period there.
- swap written words for spoken ones. "utilize" becomes "use." "in order to" becomes "to." "approximately" becomes "about."
- use contractions. "you do not need" sounds like a legal notice. "you don't need" sounds like you.
- cut the throat-clearing. "so basically what i want to talk about today is" can go. start on the thing.
- say numbers the way you'd say them. "around forty bucks" reads easier than "$39.99."
- repeat the point once. in writing, saying it twice is padding. out loud, it's how the listener catches it if they blinked.
6. end with one ask#
finish on one clear next step. subscribe, try the fix, grab the template, watch part two. one. a list of four asks gets none of them done.
how long should a video script be#
the planning number i use is about 150 spoken words per minute at an average, conversational pace. some people talk faster, a slow and deliberate read can sit well under that, and pauses eat time too. but 150 is a good place to start.
that gives you rough budgets:
- a 30 second short: about 75 words
- a 60 second short: about 150 words
- a 90 second talking head: about 225 words
- a 5 minute YouTube video: about 750 words
- a 10 minute lesson: about 1,500 words
the trap is counting words in the file instead of words you'll say. headings, notes to self and [b-roll: ...] cues all inflate the count. the speech length calculator counts only the spoken text and lets you set your own pace, so it's the quickest way to get an exact number before you film.
if your draft comes in long, cut a beat rather than squeezing every beat. a 60 second video with three solid points beats one with five rushed ones.
a worked example: an 80 second talking-head script#
here's a full script built with the steps above. it's a home office audio tip, written in markdown because that's the format open prompter reads. the headings are section markers for me, and the bracketed cue is a note for the edit.
# hook
your mic isn't the problem. your room is.
i spent money on a new microphone and my calls sounded exactly the same.
the echo was still there. here's what actually fixed it.
# why it echoes
sound bounces. it leaves your mouth, hits a hard wall,
and comes back into the mic a split second later.
bare walls, a desk, a window. every hard surface is a mirror for sound.
so a better mic just records the echo in more detail.
[b-roll: empty room, clap]
# the fix
you need soft stuff. two places matter most.
**behind you**, because that's what's facing the mic.
a bookshelf full of books works. so does a curtain or a couch.
**in front of you**, because your voice hits that wall first.
a rug on the floor and something soft on the desk side help more than you'd think.
you don't need foam panels. you need fewer bare surfaces.
# the test
here's how to check. stand where you record and clap once.
if you hear a ring after the clap, there's still work to do.
add one soft thing, clap again, and listen.
# ask
try the clap test today, before you buy anything.
and tell me in the comments what your room sounds like.
that's about 195 spoken words once you drop the headings and the cue, which lands it just under 80 seconds at an average pace. notice what it does. the hook is one line with a claim in it. the sentences are short enough to say on one breath. the bold marks the two phrases i don't want to rush. the main point gets said twice: once as "you need soft stuff" and once as "you need fewer bare surfaces."
if you'd like to know what each of those marks does on screen, the markdown syntax guide covers the four worth learning.
how to not sound like you're reading#
even a good script can come out flat. a few habits that help:
- memorize the first line. it's one sentence and you wrote it. say it to the lens, then let the prompter carry the rest.
- read slightly slower than feels natural. on camera, your normal pace sounds rushed.
- leave yourself room to drift. if a word doesn't feel right in the moment, say the one that does. the script is a guide, not a contract.
- do one full read-through before the take. you'll hear the clunky line every time, and it's cheaper to fix it in the file than in the edit.
taking your script onto a prompter#
once the script is written, reading it on camera is the easy part, if the tools stay out of the way.
open prompter reads .md files straight from a folder you pick once: iCloud Drive, an Obsidian vault, Dropbox, or anywhere in Files. you save the script there and it shows up in your library. edit the file on your Mac and the phone picks up the change. there's no paste box and no import step.
it hides the scaffolding while you read. the headings in the example show as clean section breaks, and the [b-roll: ...] cue disappears from the reading view, while your file stays exactly as you wrote it.
turn on voice tracking and the page follows what you say rather than a fixed scroll speed. the speech recognition runs on the device, and if you go quiet for about 1.5 seconds the scroll pauses, so you can stop, re-read a line, and pick back up. here's how a teleprompter that follows your voice works, and the READ and FEATHER guide shows how to tune the feel.
it also records with the front camera while you read, so for a talking-head video the phone is the prompter and the camera at once. if you put it behind a beam-splitter glass, mirror mode flips the text so it reads correctly in the reflection.
open prompter is free, needs no account, and is MIT-licensed.
common questions#
what format should a video script be in?#
for talking-head videos, a plain text file with short paragraphs works better than a two-column screenplay or a table. you're the only person reading it, and you're reading it live. markdown gives you headings and bold if you want them, and it opens in any text editor.
how many words is a 1 minute video script?#
about 150 words at an average speaking pace. if you talk fast, it can be more. if you pause a lot or speak deliberately, plan for fewer. check your draft in the speech length calculator with your own pace.
should i write every word or just bullet points?#
for your first videos, write every word. bullet points work once you're comfortable improvising on camera, but early on they lead to the "um" problem. a full script also makes your first line and your ask predictable, which are the two parts that matter most.
do i need a teleprompter to use a script?#
no. you can memorize a short script, or film it in chunks and cut between them. a prompter helps once the script is longer than you can hold in your head, because it lets you keep your eyes near the lens instead of glancing down at notes.