Short-Form Scripts: The 4-Beat Structure
Module M1 · Lesson 1.3 · MASTER (EN) — a guía / demo lesson, fully text-standalone
Course: AI Content Creation · Level: beginner (non-programmer creators)
What you'll be able to do after this lesson: take any hook from Lesson 1.2 and build a full 30-second script around it using the 4-beat structure — Hook → Build → Payoff → CTA — with delivery notes, visual-change cues, and captions planned. You'll also have an AI script generator that turns a hook + topic into a word-for-word script in seconds.
Watch the class
Primary video: The 4-beat short-form script, on screen — The Agent Academy screen-demo [own screen-demo, to be produced]
Fallback video: How I Use ChatGPT Projects to Write TikTok Hooks in Seconds — YouTube watch?v=589opy7VZTI · duration TBD [PENDING EMBED VERIFICATION]
What the video shows: the structure built live — a real hook turned into a full 30-second script, beat by beat, with visual changes and captions marked on screen.
The text is a complete standalone manual; the demo just shows the build in motion.
The concept: the hook earns attention, the script keeps it
Lesson 1.2 was about surviving the first 3 seconds. This lesson is about the other 27. A great hook with a rambling middle still dies. The 4-beat structure is the skeleton that carries the viewer to the end and, ideally, back to the start.
The four beats:
- 1. Hook (0-3s) — your scroll-stopper from Lesson 1.2. One honest promise, said and shown fast.
- 2. Build (retention) — you earn the payoff: set up the problem, add tension or steps, keep momentum. This is where most videos sag, so it's where the timing rules matter most.
- 3. Payoff — you deliver the promise the hook made. The tip, the result, the reveal. It comes before any ask.
- 4. CTA (one ask) — a single, specific action: "follow for part 2," "comment your niche," "save this." One. Two asks split attention and you get neither.
The order that matters most: payoff before CTA. New creators flip this — they ask for the follow, then give the tip. But attention is rented, not owned. You pay the viewer with value first; the ask only lands once you've proven the video was worth their time.
The three retention rules (the Build beat is where you keep people):
- Visual change every 3-5 seconds — cut to a new angle, zoom, B-roll, a prop, or a text card. The eye resets and re-engages on change.
- Pattern interrupt every 5-8 seconds — a bigger jolt than a cut: a sudden question, a sound effect, a "but here's the thing." It breaks autopilot right when attention dips.
- Captions always — most feeds autoplay muted, so a spoken-only line is invisible to those viewers (benchmark — Verified: Jul 2026). Captions are whether the video communicates at all.
The loop trick: write your final line so it feeds straight back into your first. When the video restarts (they always loop), a viewer who doesn't feel the seam watches again — and re-watch is one of the strongest signals a platform reads. Example: end on "…and that's the mistake I made too" flowing into a hook that opens "The biggest mistake beginners make is…".
How to write a 4-beat script, step by step
- Start from a finished hook (Lesson 1.2) and its one idea. The hook made a promise — your script must keep it.
- Write the Payoff second, not last. Decide the single thing the viewer walks away with. Everything else serves this.
- Fill the Build to earn that payoff — problem, tension, or 2-3 quick steps. No throat-clearing.
- Add one CTA after the payoff. One action, in your voice.
- Mark a visual change every 3-5s and a pattern interrupt every 5-8s, right in the script.
- Write the last line to loop back into the hook.
- Read it out loud with a timer. 30 seconds is roughly 70-80 spoken words. If you run long, cut the Build, never the Payoff.
- Add delivery notes — pace, emphasis, pauses. The words aren't the video; the delivery is.
Prompts (copy and use)
Prompt 1 — 30-Second Script Generator
Purpose: turn a hook + a topic into a word-for-word 30-second script with the 4 beats, visual-change cues, pattern interrupts, captions, delivery notes, and a looping last line — in your voice.
When to use it: after you've picked a hook (Lesson 1.2) and know the one idea the video delivers.
Editable variables: everything in [BRACKETS]. Paste your Brand Master Prompt (Lesson 1.1) above this so the script comes out in your voice.
[PASTE YOUR BRAND MASTER PROMPT FROM LESSON 1.1 FIRST]
Write me a word-for-word script for a 30-second vertical short.
My hook (already chosen): "[YOUR HOOK]"
The one idea / payoff the video delivers: [WHAT THE VIEWER WALKS AWAY WITH]
Platform: [TikTok / Reels / Shorts]
Structure it in 4 beats and label each one:
1. HOOK (0-3s) — use my hook above, tightened if needed.
2. BUILD — earn the payoff; keep momentum, no intro or greeting.
3. PAYOFF — deliver the promise fully, BEFORE any ask.
4. CTA — exactly ONE specific action, in my voice.
Rules:
- Total spoken words: 70-80 (about 30 seconds out loud).
- Spoken language I can say naturally on camera. No ad copy.
- Honest: deliver exactly what the hook promises, nothing inflated.
- Mark a [VISUAL CHANGE] every 3-5 seconds.
- Mark a [PATTERN INTERRUPT] every 5-8 seconds.
- Write the last line so it loops back into the hook.
- Add a one-line CAPTION cue and a short DELIVERY NOTE (pace/emphasis)
under each beat.
Variations: 3 versions with different CTAs to test asks · make the Build a 3-step list for how-to content · change the word count for 15s (35-40 words) or 45s (~110 words).
Warnings: the model loves to open with "Hey guys, welcome back" — delete any greeting; it burns your hook. It also pads the Build past 30 seconds; trim by cutting Build, never Payoff. Never publish a script you haven't said out loud.
Worked example: one hook to a full script
Hook: "Your 45-minute cardio habit is why you're still exhausted." · Payoff: 12-minute strength beats it for busy people. · Platform: Reels · ~75 words.
- Hook (0-3s): "Your 45-minute cardio habit is why you're still exhausted." —
[VISUAL: you mid-jog, then freeze-frame]; delivery flat and certain. - Build: "You're burning yourself out for a burn that barely moves the needle. More cardio isn't discipline — it's just more tired. But there's a swap." —
[VISUAL CHANGE ~5s: phone timer]·[PATTERN INTERRUPT ~8s: "But there's a swap."]; pick up pace on the swap. - Payoff: "Twelve minutes. Squats, push-ups, a plank — three rounds. You build muscle, your metabolism keeps working after you stop, and you've got energy left for your day." —
[VISUAL CHANGE: quick cuts of each move]; warm, this is the gift. - CTA: "Comment BUSY and I'll send you the full 12-minute routine." —
[VISUAL: text overlay "BUSY 👇"]; friendly, one ask, then stop.
Loop line: ending on "…routine" flows back into "Your 45-minute cardio habit…" — the viewer who wants the routine often watches twice while they comment. Captions run on every line.
Common mistakes and how to fix them
| Mistake | Fix |
|---|---|
| CTA before the payoff | Deliver the promise first. Ask once, at the end. |
| Two or three CTAs | One ask per video. Pick the single most important one. |
| A talking head that never cuts | Change the visual every 3-5s: angle, B-roll, text card, prop. |
| Sagging middle | Drop a jolt every 5-8s: a question, a "but," a sound, a hard cut. |
| No captions | Put the whole script on screen, inside the safe zone. |
| Script runs 50+ seconds | Time it out loud; cut to 70-80 words. Trim Build, keep Payoff. |
| Hard stop on the last frame | Write the last line to feed back into the hook. |
| Publishing the AI's first draft | Load your Brand Master Prompt, then edit and say it out loud. |
Exercise: write one 30-second script, word for word
- Take one finished hook from Lesson 1.2 and its one idea.
- Run Prompt 1 (with your Brand Master Prompt pasted on top) to draft the 4 beats.
- Edit it into a word-for-word script of 70-80 words, with
[VISUAL CHANGE]and[PATTERN INTERRUPT]cues marked and a delivery note under each beat. - Write a looping last line. Then say the whole thing out loud with a timer.
- Score it with the rubric (self-score, 1-5 per row):
| Criterion | 1 | 5 |
|---|---|---|
| Payoff placement | CTA comes before the value | Payoff fully delivered before the ask |
| One CTA | Two or more asks | Exactly one, in your voice |
| Visual rhythm | One static shot | A visual change every 3-5s marked |
| Pattern interrupts | None | One every 5-8s marked |
| Timing | Runs long or short of ~30s | 70-80 words, times to ~30s out loud |
| Loop | Hard stop | Last line flows back into the hook |
Expected result: one 30-second script you could shoot tomorrow — four labeled beats, cues and delivery notes in place, a loop line, scoring 24+ out of 30. If it scores under 18, the usual culprit is a CTA that jumped ahead of the payoff — move it and re-score.
Checklist before you move on
- My script has all four beats, labeled: Hook, Build, Payoff, CTA.
- The payoff is delivered before the single CTA.
- There's a visual change marked every 3-5 seconds.
- There's a pattern interrupt every 5-8 seconds.
- Captions are planned for every line (whole script on screen).
- The last line loops back into the hook.
- It's 70-80 words and I timed it out loud at ~30 seconds.
- It sounds like me talking, not like AI copy.
Toolkit for this lesson
| Tool | Link | Access |
|---|---|---|
| ChatGPT or Claude | https://chatgpt.com · https://claude.ai | Free account; Projects on paid tiers |
| Your Brand Master Prompt (Lesson 1.1) | Your saved note / Project | Free |
| CapCut | https://www.capcut.com | Free editor with auto-captions (taught in M3) |
| Phone timer / stopwatch | Built into any phone | Time your script out loud to hit ~30s |
| Google Gemini (alt) | https://gemini.google.com | Same generator prompt works; free tier |
(Verified: Jul 2026.) Paste Lesson 1.1's Brand Master Prompt above the generator; each Lesson 1.2 hook is the opening beat of a script.
Next step
You can now stop the scroll and keep the viewer to the end. You've finished Module 1 — Copywriting with AI: prompting, hooks, and scripts. Next, in Module 2 — Images with AI, you'll give these scripts a face: generate your first pro-quality visuals with GPT Image and Nano Banana, so what's on screen during every beat is as strong as the words you just wrote.
