Most YouTube scripts are written like blog posts: an intro, some sections, a conclusion. Then they get read to a camera, and the retention graph slopes down the whole way. Viewers don't read a video. They decide over and over whether to keep watching.
A script built for retention gives them a reason to stay at each of those moments. This post covers a template that does that: a clear promise up front, sections that open and close loops, pattern breaks, and visual cues written into the script so it's ready to record.
Start with an angle and a promise
Before the hook, write two sentences:
- Angle: the one reason this video is worth watching. Not "a video about home office lighting" but "your camera looks bad because of where your light is, not which light you bought."
- Promise: what the viewer will have by the end. "By the end you'll know where to put the light you already own."
The promise is a contract. Every section of the script should move toward it, and the end of the video should deliver it. If a section doesn't serve the promise, cut it or save it for another video.
The first 15 seconds: hook, then promise
The opening has one job: give the viewer a reason not to click away. A structure that works for most topics:
- Hook (first few seconds): a pattern interrupt or a claim you can back up later. Show or say something unexpected.
- Promise (by ~15 seconds): say plainly what the video delivers.
- Go. No channel intro, no "before we start, smash that like button."
Write two or three hook options and pick one. For the lighting example (made up for this post):
- "I bought a $200 light and looked worse on camera. The light wasn't the problem."
- "Move one thing on your desk and your webcam will look better. No new gear."
- [ON-SCREEN: side-by-side, same light in two positions] "Same light. Same camera. Twelve inches apart."
Only use a number like that $200 if it's true for you. If you're not sure of a figure yet, leave a placeholder (more on that below).
Open loops and payoffs
An open loop is a question the viewer wants answered that the video hasn't answered yet. A payoff is where you answer it. Retention-friendly scripts are built from loops that open and close:
- Open: "There's one position that looks fine in the mirror but terrible on camera. I'll get to it in a minute."
- Payoff: a section later, "Here's the one I mentioned: light directly behind you."
Rules that keep loops honest:
- Close every loop you open. A tease that never pays off feels like a trick, and viewers remember.
- Don't stack too many at once. One or two open loops at a time is plenty. More becomes confusing.
- Pay off sooner than feels natural. Answer the loop, then open the next one. The video should feel like it keeps moving.
A simple pattern for each section: answer the last loop → teach the next point → open a new loop.
Pattern breaks
Even a good section gets monotonous if it's one shot of a talking head for two minutes. Plan a change every so often:
- a cut to B-roll or a screen recording,
- an on-screen list or diagram,
- a quick example or story,
- a question to the viewer,
- a change of angle or location.
You don't need to improvise these in editing. Write them into the script.
Write the visuals into the script
Put visual cues on their own lines in brackets so they're easy to spot while recording and editing:
-
[B-ROLL: …]for footage that plays over your voice -
[ON-SCREEN: …]for text, graphics, or lists on screen
That turns the script into a shot list too. Your editor, or you a week later, won't have to guess what goes over each line.
Use [ADD: …] placeholders instead of guessing
When you draft a script, you'll hit facts you don't have yet: a price, a measurement, a date, a quote. Don't fill the gap with something that sounds right. Write a placeholder:
[ADD: distance from light to face in my setup]
Before recording, search the script for [ADD: and fill or cut every one. It's a small habit that keeps made-up stats out of your videos.
The template
Here's a skeleton for an 8–10 minute video. Timestamps are rough targets, not rules.
TITLE OPTIONS (3) + THUMBNAIL TEXT (2-4 words)
0:00 HOOK
[ON-SCREEN: ...]
Line that interrupts the pattern or makes a backable claim.
0:10 PROMISE
"By the end of this video you'll ..."
0:20 SECTION 1: first point
Payoff of the hook's implied question.
[B-ROLL: ...]
Open loop: "... I'll show you why in a minute."
2:30 SECTION 2: second point
Pay off the loop from section 1.
[ON-SCREEN: list / diagram]
Pattern break: example or question to viewer.
Open loop for section 3.
5:00 SECTION 3: biggest point
Pay off. [ADD: fact you still need to check]
7:30 RECAP: deliver the promise in one or two lines
8:00 CTA + END SCREEN
One natural ask, then tease the next video:
"If X was your problem, the next video covers Y."
A worked snippet
Here's what the start of the lighting video might look like in that format. It's an example written for this post.
0:00 [ON-SCREEN: same face, two shots side by side]
"Same light. Same camera. Twelve inches apart."
0:05 "If you've bought a light and still look washed out,
the light's probably fine. It's where it is."
0:12 "By the end, you'll know exactly where to put the light you own."
0:20 [B-ROLL: light on the desk, aimed straight at face]
"Most people put it right in front of them, straight on.
That's the first thing we'll fix, but it's not the worst
position. The worst one looks fine in the mirror.
I'll get to it."
End screen: re-hook, don't sign off
"Thanks for watching, see you next time" is where people leave. End with one natural ask and a reason to watch something else: name the next video and why it's the logical next step. Write the title and thumbnail text alongside the hook, too, since they make the same promise from outside the video. Write a few options so you have something to A/B test later.
Automating the first draft with YouTube Script Writer
Getting angle, promise, hooks, loops, cues and placeholders right on every video takes time. My team built an AI agent for that first draft called YouTube Script Writer. It runs in a chat on Capafy, an AI agent marketplace.
What you send: a topic, a draft, or a product.
What you get back, according to the listing:
- 3 title options and a thumbnail-text idea,
- 3 hook options for the first 15 seconds, followed by the video's promise,
- a timestamped script where each section opens and pays off a loop, with
[B-ROLL: …]and[ON-SCREEN: …]cues, - a CTA and end screen that teases the next video,
-
notes on pacing and what to A/B, plus any
[ADD: …]facts you need to fill in.
The listing says it writes only from what you provide, doesn't invent facts or stats, and makes no promises about views or subscribers. You still record and edit, and you should read the draft as you'd read your own: check that it sounds like you and fill every placeholder before you hit record.
Pricing on the listing starts at $9.99 (weekly or monthly plan), with a yearly plan at $99.99.
Wrap-up
- Write the angle and the promise before the hook.
- Hook, then promise, within the first 15 seconds. No intro.
- Build sections as loops: pay off, teach, open the next.
- Plan pattern breaks and put
[B-ROLL]/[ON-SCREEN]cues in the script. - Use
[ADD: …]instead of guessing facts. - End by pointing to the next video, not by signing off.
Disclosure: my team built YouTube Script Writer and sells it on Capafy.
Top comments (0)