Video Script Prompts: Hook, Structure and CTA in One Pass

Ask ChatGPT for "a video script about my topic" and you usually get 800 flat words that sound like a school essay. The prompt is not the culprit. It never locked anything down before writing. The difference between a script people watch to the end and one they scroll past lives in the fields you fill before you ask for the first line: platform, duration, audience, promise and CTA. A good video script prompt works in one pass when those variables are fixed first. This guide shows you how to get hook, structure and CTA from a single generation, and ends with three ready-made prompts that already do it.

Lock the Variables Before You Ask for Lines

Video script prompts work when you lock platform, length, audience and offer before you ask for a single line. Leave them open and the model fills the gaps with generic defaults: a narrator voice, no clear viewer, a vague ending. Kits for hooks, B-roll cues and CTAs only help after that brief is concrete. Define the goal first, then the format, then the words. That sequence separates a script you can film tomorrow from a draft you will rewrite twice.

The Five Fields That Change Everything

Five fields change every script you generate: platform (long-form YouTube, Shorts or TikTok), target duration in minutes, audience (who is watching and what they already know), the promise (what they get if they stay) and the CTA (the single action to take at the end). Type them as plain lines and the model writes inside them. Skip one, say the audience, and the tone drifts back to generic creator-speak.

One Script, Different Dialects

The same topic reads differently on every platform. A ten-minute video can take its time: context, stories, a slower build. A 45-second Short carries one idea, one hook and one payoff. Asking for the right format is part of the prompt itself. If you do not name it, the model picks one for you, usually the essay.

A Video Script Prompt Needs Three Parts to Hold Retention

Every high-retention video follows a three-part structure: a hook that earns the first 30 seconds, a body that delivers value in digestible segments with pattern interrupts every two to three minutes, and a close that drives the next action. When you write a video script prompt, ask for those three blocks by name. Models respond well to named containers: hook, body, close. They respond poorly to "make it engaging", which returns adjectives instead of architecture.

Hook That Earns the First 30 Seconds

In long-form, the first 30 seconds decide whether the rest of the video exists for that viewer. The pattern that keeps working is a pattern interrupt plus a promise of payoff: a phrase that breaks the scroll, followed by what they win if they stay. A hook that only surprises is a trick. A hook that surprises and promises is a contract.

Body With Rehooks, Not Walls of Text

A retention script builds momentum with the promise, pacing shifts, open loops, rehooks, examples, the payoff and a reason not to leave yet. Name those elements in the prompt and the model spaces them out instead of dumping paragraphs. Open loops are cheap to request: end a segment with a question the next one answers.

Close That Drives the Next Action

The close restates the payoff and leaves exactly one CTA. Two CTAs compete and neither wins. The close is where the promise you opened with gets collected.

Short-Form Scripts Are Screenplay Notes, Not Essays

On TikTok and Shorts the hook lives in the first one to three seconds: one spoken line plus an optional visual direction. The script is closer to a shooting note than finished copy, because verbal delivery, on-screen movement and editing pace complete around 60 percent of what the script sets up. Write short-form scripts as fragments with [VISUAL] cues, not paragraphs. The viewer never reads them. The camera and the editor do.

Loop and Payoff

Short-form is written for the loop. The ending should invite a replay or a comment, and the payoff should land seconds after the hook promised it, not minutes later. If the last line echoes the first, you get the repeat view without asking for it.

Give the Script a Measurable Length

Natural speaking pace for on-camera video sits around 130 words per minute, and a high-energy TikTok presenter runs closer to 160. That difference matters when you prompt. Ask for "target 8 minutes, about 1040 words" and the model adjusts the number of beats to the real duration instead of handing you a 20-minute script when you asked for eight. Duration is a constraint like audience: without it, the model optimizes for completeness, not time.

Ask for Timestamps

Request a timestamp for every section: 0:00-0:30 hook, 0:30-2:00 beat one, and so on. Timestamps force the model to size each block against the total runtime and give you an editing index for free.

Make It Sound Like a Person, Not a Narrator

An AI script announces itself when it sounds like a generic narrator. Before generating, run the concept through three tests: does it have a unique angle, does it resolve the curiosity it opens with, and does it sound like one specific person talking? If any test fails, fix the prompt, not the output. The model will not add a point of view that was never requested.

Rules That Kill the Robot Tone

Add explicit style rules to the prompt: conversational spoken English, short sentences, contractions, [VISUAL] markers for B-roll, two rhetorical questions and one moment of tension. Models follow named rules far better than "sound natural", and a checklist changes the tone on the first try.

Fill in the Voice, Not Just the Facts

"Sounds robotic" is almost always a facts-only prompt with no voice. Define the voice in one line: direct, dry humor, enthusiastic. That single line moves the result more than any generic style instruction, because it gives the model a person to imitate instead of a register to apply.

Three Prompts That Already Did the Work

The pattern in this post is not theoretical. Three prompts in the LikePrompts catalog fix the fields, structure in blocks and close with one CTA. They run the same logic for adjacent jobs.

Package the Video Before You Film It

Titles and thumbnails are the promise your hook has to keep. The YouTube Title and Thumbnail Copy Lab takes your video topic, channel, niche and subscriber count and returns ten titles by formula, the top three ranked by estimated CTR, thumbnail overlays and SEO copy in one pass. Use it before you write a word of the script so the hook and the packaging point in the same direction.

Turn One Script Into Many Pieces

The repurposing pattern for video is the same one podcasters use: one long asset becomes notes, clips and captions. The Podcast Show Notes and Clips prompt turns a transcript into episode titles, timestamped notes and quotable clips with captions. Paste your finished script into it and the output is a distribution plan, not just a summary.

Same Hook-CTA Pattern for Social Copy

The skeleton of a short script also works as social copy: hook in the first line, value, social proof, one CTA. The Instagram caption generator builds captions that way, with hashtags on top. When the video is live, that prompt turns the same angle into the captions that distribute it.

Your move is simple: fill the five fields, name the three parts, and let one generation do the work. Lock the promise first with the YouTube Title and Thumbnail Copy Lab, then apply the same hook-and-close logic to every asset that comes out of the video. The prompt handles the structure. You bring the video.