How to Write a YouTube Video Script (Template + Examples)

Table of contents
  1. The hook: your first 3 to 10 seconds
  2. The retention structure
  3. Writing for the ear, not the page
  4. B-roll, on-screen text and other cues
  5. Where the CTA goes
  6. A full sample script
  7. Shorts, Reels and TikTok are a different script
  8. A template you can reuse
  9. Using AI without sounding like everyone else
  10. Write the next one

Most videos are not lost at minute six. They are lost in the first ten seconds, and everything after that is a slow leak. This guide gives you a hook you can write in one line, a structure that keeps people watching, the cue system editors actually want, and a complete sample script you can copy.

The hook: your first 3 to 10 seconds

The opening has one job — make the next sentence feel necessary. It is not an introduction, it is a promise or a problem.

Three things to cut from every opening you write: your name, the word "welcome", and any request to like and subscribe. All three can wait.

Five hooks you can steal the shape of

  • The wrong assumption. "You have been proving your sourdough in the coldest room in your house. That is why it tastes flat."
  • The cost. "I ran this campaign for six weeks before I noticed the setting that was quietly wasting most of the budget. Here it is."
  • The visible result first. "This is the shelf. That is the same shelf four hours earlier. Two tools, no workshop."
  • The fair fight. "Same recipe, same oven, one uses baking soda and one uses baking powder. One of them is going to be a biscuit and one is going to be a hockey puck."
  • The blunt question with a real answer. "Is a mirrorless camera worth it for a beginner? No — and I will show you the three situations where that flips."

Every one of those states a specific thing and leaves a gap. Compare that to "Hey guys, today we are going to talk about bread." Nothing is at stake, so nobody stays.

The retention structure

Under the hook sits a repeatable shape. It works for a three-minute explainer and a twenty-minute deep dive; only the number of sections changes.

SectionShare of runtimeJob
Hook3–10 secondsName the problem or promise
Stakes10–20 secondsWhy it matters to this viewer, now
Roadmap5–10 secondsOne line: what they will have by the end
Body60–70%Three to five sections, each a small payoff
Payoff10%The result, the demo, the answer
CTA and next video10–15 secondsOne ask, one link

Two habits do most of the work inside the body. First, open a loop and close it later: "the third mistake is the expensive one — I will come back to it." Second, start every section with its own mini-hook rather than a number: not "step two", but "here is where most people stop too early".

Write a visual change every fifteen to twenty seconds, even if it is only a cut to a different angle. A talking head held for ninety unbroken seconds is where viewers leave.

Writing for the ear, not the page

A script that reads well often sounds like a press release. Fixes that take minutes:

  • One idea per sentence. If you need a comma to hold it together, split it.
  • Contractions everywhere. "You'll", "it's", "here's".
  • Numbers said the way you would say them: "about a fiver", not "approximately GBP 5.00".
  • Second person. "You" far more often than "we" or "one".
  • Read it aloud at pace. Anywhere you stumble, the sentence is wrong — not your mouth.

A rough timing rule: most presenters land near 150 words a minute. So a 400-word script is roughly two and a half to three minutes on screen. Count words before you book a studio.

B-roll, on-screen text and other cues

Write your script in two columns, or tag lines inline. What matters is that the editor never has to guess.

  • [B-ROLL] — what the viewer sees while you talk. Be concrete: "hands scraping dough off the counter", not "baking footage".
  • [TEXT] — words on screen. Three to five words, never a full sentence.
  • [SFX] — a sound effect, used sparingly so it still means something.
  • [CUT] — a jump cut or angle change. Mark these while writing; you will write tighter.
  • [GRAPHIC] — a chart or diagram, with the one number you want read.

Rule of thumb: never put on-screen text that duplicates what you are saying word for word. Use it for the label, the number, or the joke.

Where the CTA goes

Not at the start. One soft moment in the middle, right after a payoff — "if that fixed it for you, subscribing helps" — and one clear ask at the end. Never two asks at once. If you want the comment, do not also want the subscribe and the link.

End on the next video, not on "thanks for watching". The final line should point somewhere: "the proving box I mentioned is in the next video — it cost about the price of a takeaway."

A full sample script

Here is a complete short explainer, roughly two and a half minutes. Cues included.

TITLE: Why Your Bread Is Flat (It's Not the Yeast)

[HOOK — 0:00]
[B-ROLL: flat loaf pulled from oven, slow]
You've been proving your dough in the coldest room in the house. That's why it comes out like this.
[TEXT: Not the yeast]

[STAKES — 0:08]
[CUT to presenter]
Everyone blames the yeast, buys new yeast, gets the same loaf. The yeast is almost never the problem. Temperature is.

[ROADMAP — 0:18]
Three things to check, in order of how much they cost you. The third one is the one nobody checks.

[SECTION 1 — 0:25]
[TEXT: 1. Where it proves]
Your kitchen is not one temperature. The counter by the window is cold. Above the fridge is warm.
[B-ROLL: hand thermometer on counter, then above fridge]
Six degrees of difference in the same room. That's the difference between a two-hour prove and a five-hour one.

[SECTION 2 — 0:55]
[TEXT: 2. How you know it's ready]
Stop using the clock. Use the dough.
[B-ROLL: finger poke, dough springs back slowly]
Press two knuckles in. If it springs back fast, wait. If it stays dented, you've gone too far. Slow and halfway back — that's the one.

[SOFT CTA — 1:20]
If that poke test is new to you, subscribing is the easiest thank you.

[SECTION 3 — 1:28]
[TEXT: 3. The expensive one]
Here's the one I promised. Your oven lies.
[GRAPHIC: dial says 220, thermometer says 196]
Most home ovens run cold, and mine was off by more than twenty degrees. A three-pound oven thermometer fixed two years of flat loaves.
[SFX: oven door]

[PAYOFF — 2:00]
[B-ROLL: same recipe, risen loaf, cut open]
Same flour. Same yeast. Warm spot, poke test, corrected oven.

[CTA — 2:15]
Next video is the proving box I built for about the price of a takeaway. It's on screen now.
[TEXT: Proving box →]

That script is around 380 words. If yours is 900 words and you promised a two-minute video, the problem is the script, not the edit.

Shorts, Reels and TikTok are a different script

Vertical short-form is not a trimmed long video. The differences are structural:

  • No roadmap. There is no time to say what you are about to do. Just do it.
  • The hook is the first second, not the first ten. Often the first frame — the result on screen before you say a word.
  • One idea only. Three tips becomes three videos.
  • Write for silent viewing. Assume captions carry the message; make the visual readable with the sound off.
  • Loop the ending. Close on a line that makes the opening frame make more sense, so a rewatch feels intentional.
  • The CTA is usually the comment. A question beats a subscribe ask in short form.

Same idea, two scripts: "Three reasons your bread is flat" for YouTube, and "Your oven is lying to you" for Shorts. Do not try to make one file serve both.

A template you can reuse

Working title: [the promise, in the viewer's words]
One thing they'll be able to do: [single outcome]
Hook (1 line): [wrong assumption / cost / result first / fair fight / blunt question]
Stakes (1–2 lines): [why this matters now]
Roadmap (1 line): [what's coming, with an open loop]
Section 1: mini-hook → point → [B-ROLL] → [TEXT]
Section 2: mini-hook → point → [B-ROLL] → [TEXT]
Section 3 (close the loop): mini-hook → point → [GRAPHIC]
Payoff: [the result on screen]
CTA: [one ask] + [next video]
Word budget: [minutes × 150]

Fill the hook and the payoff first. If those two lines are not interesting to you, no amount of middle will save it.

Using AI without sounding like everyone else

Drafting is where a tool earns its place. Ask for ten hooks and keep one. Ask for the section order twice and compare. Our AI video script generator is free, needs no sign-up and streams as it writes, so you can kill a bad opening after two lines instead of reading the whole thing.

Three prompts worth keeping:

  • "Ten hooks under 15 words for a video about oven temperature and flat bread. No questions that start with 'have you ever'."
  • "Turn this outline into a 400-word spoken script. Second person, contractions, one idea per sentence, no filler intro."
  • "Rewrite this section for the ear: shorter sentences, cut anything I'd never say out loud."

Then edit hard. The model gives you a competent middle; your specifics — the twenty-degree oven, the three-pound thermometer — are the reason anyone watches. For scripted formats beyond talking head, our AI script generator handles sketches and dialogue scenes, and if you need story shape rather than structure, start with the free AI story generator. It also helps to think in visual storytelling terms: what does the viewer see while you talk?

Write the next one

Good scripts are mostly deletion. Write the hook, write the payoff, connect them with three sections, then cut a third of the words and read it aloud.

Draft yours now with the free YouTube script generator, or start from an idea with our AI story generator.

Keep reading

You might also like

More Guide