How to write an AI video prompt, with examples

What should an AI video prompt include?

An AI video prompt should tell the tool what it cannot guess: what the film is for, who it speaks to, the one point it must make, the feeling and the ending. Voom’s studio reads every prompt into seven decisions before it designs anything, so the best prompts answer all seven.

These decisions are the first step in Voom’s studio for AI motion graphics. The look, script, voice, music and length all follow from them.

Decision	What the studio decides	What to write
1. Kind	Product film (it launches, promotes, demos or sells a product, app, service or company) or explainer (it teaches a topic, idea or process); a website link usually means a product film	Launch, feature announcement, explainer or ad
2. Name	The product or brand name, or a short topic title	The name or the topic
3. Promise	The one line the whole film proves, in the viewer’s outcome language, not a feature list	What changes for the viewer
4. Audience	Who the film speaks to	One specific group
5. Call to action	The single action at the end; none for a pure explainer if none fits	One action
6. Mood	3 to 6 words for the look, the music and the voice	Two to four tone words
7. Length	On Auto, what the content deserves, up to 45 seconds: about 15 to 25 seconds for a tight teaser, 25 to 40 for a feature launch, 30 to 45 for an explainer, never padded	Nothing, if you set a length

Bottom line: a prompt that states the kind, name, promise, audience, mood and ending settles six decisions. Length belongs in the options.

What formula covers all seven decisions?

The formula puts the decisions in the order a viewer meets them: what the film is, who it is for, the point, the feeling and the ending. Fill the brackets and stop at three sentences; a longer prompt usually hides a second message.

[Kind of film] for [name], [what it is in a few words]. For [audience]. The point: [the promise, as the viewer’s outcome]. [Two to four mood words]. End on [one action].

• The point is the outcome, such as “bills settled before payday”, not the features behind it.
• One action gives the end card its line, shown with the logo for the last 2.5 to 3.5 seconds.

Google’s advice for video ads ends the same way: include a call to action and say what you want viewers to do (Google Ads Help). One action is easier to say and to show than three.

How do mood words change the voice and the script?

Mood words set more than the look. Voom’s studio picks 3 to 6 of them from your prompt, and they also choose the narrator’s pace and how many words the script may use: calm words give a relaxed read and a shorter script, and energetic words a brisk read and a fuller one.

Mood words	Narrator’s pace	Script budget
calm, slow, gentle, serene, soft, luxurious	Relaxed	1.9 words a second
premium, elegant, cinematic, confident, precise, minimal	Balanced	1.9 words a second
bold, energetic, punchy, fast, hype, playful	Brisk	2.5 words a second
Others, such as friendly, warm or clear	Balanced	2.2 words a second

When a mood mixes words from the first and third rows, the first row wins, so “calm but punchy” gets a relaxed read and the 1.9-word budget. With the voice on Auto, the mood also casts the narrator, and the original score follows the same mood.

A 30-second script gets 48 words with calm mood words, 56 with friendly and 63 with energetic. Our guide to how long a product video should be lists budgets from 15 to 90 seconds, and the product launch video script templates turn them into lines.

AI video prompt examples for four kinds of film

The four prompts below are hypothetical: they show the formula at work, not films that exist. Two are sample prompts from Voom’s own pages and two were written for this guide, and each note says what the prompt decides and what it leaves open.

A product launch film

Launch video for our budgeting app for couples. Confident and warm, ending on our website address.

This sample prompt covers the kind, audience, mood and call to action, but leaves the name and the promise to the studio. “Confident” points the script to the 1.9-word budget: 48 words at 30 seconds. A stronger version adds the missing two:

Launch video for [app name], a budgeting app for couples. For couples who share bills. The point: you both see every bill and plan the month together. Confident and warm. End on our website address.

An explainer

Explain how a heat pump keeps a home warm in winter by moving heat instead of making it. Friendly and clear, for homeowners.

This sample prompt covers almost every decision: “explain” sets the kind, “moving heat instead of making it” is the promise and homeowners are the audience. “Friendly and clear” points to a balanced read at 2.2 words a second, and a pure explainer needs no call to action.

A social media ad

Ad for [studio name], a pottery studio with weekend classes for beginners. For adults who want a hobby away from screens. The point: you leave with a bowl you made yourself. Playful and bold. End on: book a Saturday class.

Pair this prompt with 9:16, 15 seconds and captions on. “Playful and bold” points to a brisk read at 2.5 words a second, or 30 words in 15 seconds. Short suits paid feeds: X recommends video ads of 15 seconds or less (X Business), and LinkedIn 15 to 30 seconds (LinkedIn Help).

A film made from a product link

A film for this page, for owners of small bakeries. The point: online orders arrive in one list, ready to bake. Calm and precise. End on: start a free trial.

Paste the product page’s link with this prompt: Voom reads the site’s colours, fonts, copy, screenshots and logo, so the link supplies the name and the look. “Calm and precise” points to a relaxed read on the 1.9-word budget. See also how to turn a website into a promo video.

What should you leave out of a video prompt?

Leave out what the studio takes better from the source and anything that gives the film a second job: colours and fonts when you add a link, shot lists, long feature lists and figures you cannot source. Each one repeats the page or dilutes the promise.

• Colours and fonts with a link. The site supplies them; without one, Voom designs the look from the topic and your mood words.
• Shot lists. Describe a shot only when it must appear, such as a real product screen.
• Every feature. Name the two or three that prove the promise.
• Unsourced numbers. A product film states only figures its sources state, so give each figure’s source.
• A length you already chose. The studio plans to the length in the options.

How do you write a video prompt, step by step?

Write the prompt in the order of the seven decisions, then set the rest in the options. The five steps work for any film, from a 15-second ad to a one-minute explainer, and each settles a decision the studio would otherwise make alone.

1. Name the kind of film. Launch, feature announcement, explainer or ad: the kind sets the story’s structure.
2. Write the promise in one sentence. Say what changes for the viewer, in their words.
3. Name one audience. A specific group gets examples, a voice and music that fit it.
4. Pick two to four mood words in one direction. Calm words slow the read; energetic words speed it up.
5. End on one action. Name the one thing viewers should do, or leave it out for a pure explainer.

Then set the size, the length or Auto, the voice and captions, and add a link if the film is about your product. See how Voom works for every stage from prompt to MP4.

Common mistakes in AI video prompts

Most weak prompts fail in one of five ways, each tied to one of the seven decisions. They list features instead of a promise, mix moods, ask for several endings, quote unsourced numbers or describe camera footage a motion graphics tool does not make.

• Features instead of a promise. Fix: one sentence on the viewer’s outcome.
• Mixed moods. “Calm, bold and playful” pulls the read two ways. Fix: choose one direction.
• Several calls to action. Fix: one action.
• Numbers without a source. Fix: source every figure.
• A text-to-video prompt. “A woman walks through a sunlit kitchen” describes filmed footage. Fix: describe the message, audience and mood; motion graphics are built from type, shapes, charts and screens.

When is Voom the wrong tool for your prompt?

Voom is the wrong tool when the prompt needs what motion graphics cannot show: real people or places on camera, a presenter talking to the lens, characters who act or detailed 3D. Those belong with a film crew, an animator or a 3D artist, and a click-by-click tutorial with a screen recorder.

Length and control are the other limits. Voom makes films of up to 30 seconds on Free, 90 on Pro and 180 on Business, and a finished film cannot be edited, only remade as a variation. If every frame needs sign-off, hire a motion designer.