Your launch video script, structured
The beat structure every good launch video follows: hook, problem, feature beats, close. Timed to your length and filled with your product.
the structure
How to write a launch video script
A launch video script has four beats: a hook in the first three seconds, the problem in one sentence, two or three feature beats shown rather than listed, and a close with a single call to action. The generator above drafts all four with timings, visual cues and voiceover lines.
- Hook (0–3s) — the result or the claim, before any logo.
- Problem (3–10s) — one sentence the viewer already agrees with.
- Feature beats (10–60s) — show each one working; no bullet lists on screen.
- Close (last 5–10s) — one call to action, spoken and on screen.
timing
How many words fit in your runtime
Natural voiceover runs at roughly 150 words a minute. Scripts fail in the edit because they were written to a page count instead of a clock, so fix the word budget before the first sentence.
15 seconds — about 38 words
A bumper. One claim and one visual, with the product doing the thing for most of the runtime. Works as a teaser before launch day or as a paid pre-roll where the viewer can skip at five seconds.
30 seconds — about 75 words
One hook, one feature, one close. There is no room for a problem statement, so the hook has to imply it. This is the social-feed length, and the hardest on this list to write.
45 seconds — about 112 words
Buys back the problem statement without stretching the middle. A good default when the hook needs a sentence of context before the feature makes sense.
60 seconds — about 150 words
Hook, problem, two feature beats, close. The default for a Product Hunt or launch-day video, and long enough that each beat gets a visual of its own rather than a shared montage.
90 seconds — about 225 words
Adds a third feature beat or a proof moment — a metric, a customer line, a before-and-after. Past this length feed viewers drop off, so the extra 30 seconds should earn its place with evidence, not more features.
2 minutes and beyond
Documentation territory rather than launch territory. Works for a walkthrough someone chose to watch from your docs or a sales follow-up, and fails as a hero video on a landing page.
what goes wrong
Four ways launch scripts fail
Opening with the company. A logo animation and an “introducing” spends the only three seconds you are guaranteed on information the viewer cannot use. Lead with the outcome and let the name arrive once someone wants it.
Listing features instead of showing one working. A bulleted screen of six capabilities is faster to write and slower to believe. One feature completing a real task is the version a viewer can repeat to a colleague.
Writing for sound in a muted world. Most feed video plays without audio, so any point carried only by the voiceover is a point most viewers never receive. Every beat needs a visual or an on-screen line that survives silence.
Closing with several asks. Sign up, star the repo, follow us, join the Discord. Four calls to action perform worse than one, because a viewer deciding between options usually picks none. Choose the single next step that matters this week.
the hook
Six openings that survive a feed
The hook is the only part of the script most viewers will ever see, and it is the part most often written last and worst. These six patterns cover nearly every product worth launching. Pick the one your product can back with a real number or a real demonstration, because a hook that overpromises loses the viewer at the first beat that fails to deliver.
The result, up front
Open on the finished output, then reveal how long it took. Works when the output is visually impressive and the time is surprising. Fails when the result needs explaining before it lands. If a viewer cannot tell what they are looking at in one second, this is the wrong opening.
The before and after
Two states, cut hard between them, no narration needed. The strongest option for anything that replaces a visibly painful process, because the viewer does the arithmetic themselves and believes their own conclusion more than your claim.
The specific number
“Forty minutes to four.” A number is concrete, checkable and repeatable, which means a viewer can quote it to a colleague. Use one you can defend, and show the measurement somewhere in the video rather than asserting it.
The named frustration
Say the thing your buyer complains about, in their words, in one sentence. Effective when the pain is universal in your category and nobody else is naming it plainly. Risky if the frustration is niche, because everyone else scrolls.
The unexpected demonstration
Do the thing that seems like it should not work, on camera, in one take. High trust, high effort, and impossible to fake, which is why it converts when the product can take it.
The direct question
Ask what the viewer is already thinking. Weakest of the six on its own, because questions invite a mental “no” and a scroll, but it works when the answer arrives inside two seconds and contradicts what they expected.
the muted problem
Writing for a video nobody hears
Most feed video plays without sound, and a meaningful share of viewers never turn it on. A script written as narration with visuals attached loses everything on the way to those people. The fix is to decide, beat by beat, which channel carries the point.
Work through the script twice. On the first pass, read only the voiceover and ask whether it makes sense alone; that is the audio experience for someone listening while doing something else. On the second, cover the voiceover and look only at the visual cues and on-screen text; that is what most of your audience actually gets. Any beat that fails the second pass needs its point moved onto the screen, either as a caption, a title, or a demonstration that speaks for itself.
This changes how you write the lines themselves. Numbers, product names and any claim you want repeated belong on screen, because a spoken figure is gone in a second and a written one can be screenshotted. Narration is then free to do what text is bad at: pacing, tone, and the connective sentences that would clutter a frame. The two channels should carry different weight instead of duplicating each other word for word, which is the most common failure in scripts written as documents first and storyboarded later.
handing it over
Briefing a human editor versus an AI agent
The same script gets used differently depending on who builds the video. A human editor reads intent, so an instruction like “make this feel fast” gets translated through experience into cuts, pacing and music. They will also tell you when a beat is not working, which is a service you are paying for.
An agent reads structure. It follows a beat sheet with explicit time bands, one visual cue per beat and the exact on-screen text far more reliably than it interprets a mood. Where a human fills gaps with judgement, an agent fills them with a guess, so the specificity that would feel patronising in a brief to a person is what makes an agent land it on the first pass. Write the numbers, name the elements, and say what appears on screen rather than describing how it should feel.
The advantage arrives at revision time. A human revision costs a round trip and a fee; an agent revision costs a sentence, because the project stays editable and re-renders locally. That changes what you commit to in the first draft. Shipping a version and cutting two beats after watching it is cheaper than arguing about them in a document, which is the opposite of how scripts get approved when every change has a price.
the close
Ending on something a viewer can act on
Closes fail more quietly than hooks. Nobody reports that a video ended badly; the click does not happen, and the video gets blamed for the wrong reason. Three things decide whether a close works.
The ask has to be one thing, and it has to be the thing you actually want this week. Sign up, star the repo, join the waitlist, book a demo — each is reasonable, and listing all four is how you get none. A viewer weighing options defaults to closing the tab, so the choice you present should already be made.
It has to be readable without sound and without a link. In most places a video plays, the URL is not clickable, so a spoken “link in bio” over a logo leaves the viewer with nothing to type. Put the destination on screen as text, and keep it short enough to remember for the two seconds it takes to switch apps.
And it has to hold. A close that cuts to black the instant the narration stops gives the viewer no time to read what you just put on screen. Hold the final frame for two or three seconds. On a looping platform this is doubly true, because the held end state is what a second-time viewer sees first, and it is the only frame that appears twice.
Cut the outro animation. A three-second branded sting after the call to action moves the ask away from the end of the video, where attention is highest. Put the logo in the same frame as the ask, and end there.
Write the close before the middle. It is the only beat with a job that can be stated in a sentence, and knowing where the video is going changes which features earn a beat. A script written front to back tends to accumulate everything the team wanted mentioned and then discover, at the end, that there is no room left for the ask. Working backwards from the single action you want makes the cuts obvious. Anything that does not move a viewer toward that action goes first, and the argument about which feature to drop stops being a matter of taste.
CoAnimator is the full studio, driven by your AI agent.
- Real-footage effects
- Motion graphics
- 3D scenes
- Voiceover
- Unlimited local renders
FAQ
How do I write a launch video script?
Work in beats, not paragraphs: a hook in the first three seconds, the problem in one line, two or three feature beats with a visual cue each, and a close with a single ask. Time each beat before writing copy — a 60-second video holds roughly 150 spoken words. The generator above scaffolds exactly this structure around your product.
Is this launch video script generator free?
Yes — free in the browser, no account, and the output is yours to use anywhere. It generates a timed beat sheet with bracketed visual cues and voiceover lines, formatted so you can paste it straight to an AI agent or a human editor.
What makes a good launch video hook?
A claim or contrast the viewer can evaluate instantly: the outcome your product creates, the pain it deletes, or a before-and-after in one sentence. Avoid opening with your logo or “introducing” — platforms count the skips before your name lands. The strongest hooks are specific: a number, a time saved, a thing made possible.
How long should a launch video script be?
Aim for 45 to 90 seconds: long enough for a hook, problem, three feature beats, and a close, short enough to survive feed attention. At a natural voiceover pace of about 150 words per minute, that’s 110 to 220 words of copy. Cut features before cutting the hook or the close.
What do I do with the script once it’s written?
Hand it to whoever builds the video. In CoAnimator, you paste the beat sheet to your agent, which turns each beat into an animated scene, generates the narration through the TTS plugin, and renders the finished MP4 on your machine. The script’s time bands become the timeline.
Why this script structure?
Hook before product, problem before solution, one feature per beat, close with the ask — the structure that survives on every platform.
Can it write the copy for me?
It scaffolds honest placeholder lines around your inputs; the voice is yours to sharpen. Inside CoAnimator, your agent drafts and revises the full narration with you.