AI Video Ad Prompt Examples: Turn a Creative Brief into Clear Scenes

An AI video prompt is not a replacement for an advertising brief. It is a scene instruction. The strongest prompts begin after a team has decided the product facts, audience moment, visual purpose, and call to action. Without those decisions, adding more adjectives often produces more variation, not more clarity.

The examples below show how to convert a reviewable creative brief into scene-sized instructions. They are not claims that a particular model will produce a particular result. Replace bracketed details with assets and facts you are allowed to use, then inspect every output before publication.

UGC Labo helps prepare the creative brief that comes before prompting. It does not render, export, or deliver a finished video.

Start with a brief, not a long prompt

Before writing a scene, record five decisions:

1. Product truth: exact product, variant, packaging, visible details, and approved claims.
2. Audience moment: who is watching and what situation makes the product relevant.
3. Scene job: hook, context, demonstration, proof, or CTA.
4. Visual constraints: aspect ratio, product placement, camera movement, readable text space, and supplied reference assets.
5. Review boundary: what must not change, what needs human approval, and what claims must not be introduced.

The AI UGC video planning checklist can help create these inputs. For timing and spoken lines, use the 15- and 30-second script examples.

A practical prompt structure

Use one prompt per scene. A reliable structure is:

Scene purpose: [hook / product use / detail / CTA]
Subject: [product and person, if any]
Action: [one visible, accurate action]
Setting: [specific ordinary context]
Composition: [vertical or horizontal frame, subject placement]
Camera: [one simple movement or static shot]
Product accuracy: [details that must remain unchanged]
On-screen text space: [where a designer can add verified copy]
Avoid: [unsupported claim, distorted labels, extra products, unreadable text]

Keep the language concrete. “A modern, beautiful, viral ad” gives little direction. “Vertical close-up of the approved [product] on a bathroom counter, the label facing camera, hand opens the cap, clean space in upper third for caption” describes a scene that can be reviewed.

Example 1: product demonstration prompt

Use this for a short scene that shows a real product action.

Scene purpose: Demonstrate how [product] is used during [specific routine].
Subject: The approved [product name and variant] only; packaging and label match the supplied reference.
Action: A hand performs [accurate action] once, at a natural pace.
Setting: Bright [kitchen / desk / bathroom] with a simple, uncluttered background.
Composition: Vertical 9:16 frame; product remains centered in the lower half.
Camera: One steady close-to-medium shot with a slow push-in.
Product accuracy: Preserve [color, cap, label, logo, included component].
On-screen text space: Leave the upper third uncluttered for approved captions.
Avoid: Changing the label, inventing accessories, showing an unapproved result, or adding text inside the generated image.

This prompt asks for one action and one product state. Add captions in the editing step rather than relying on generated text to be accurate.

Example 2: creator-style context prompt

Creator-style does not have to mean a false testimonial. The scene can present a product use case without claiming the person is a real customer.

Scene purpose: Show the viewing moment before introducing [product].
Subject: A generic adult presenter preparing for [ordinary routine]; no identity, employer, expertise, or customer status is implied.
Action: The presenter places the approved [product] beside items used in [routine].
Setting: A realistic home [desk / entryway / countertop] in daylight.
Composition: Vertical 9:16, medium shot; product is visible but not obscured.
Camera: Static handheld-style framing with natural, minimal movement.
Product accuracy: Use the supplied product reference and preserve its visible details.
On-screen text space: Keep the top area clear for the line “[approved use case].”
Avoid: Review-style gestures, before-and-after imagery, fabricated results, medical or professional cues, and altered packaging.

Pair this scene with neutral copy such as “Designed for [approved use case]” rather than “This changed my routine.” If you use a voiceover, ensure the wording is a brand statement, not a simulated personal experience.

Example 3: product-photo-to-video prompt

When the source is a still image, protect the product identity before asking for motion. The product photo to ad video guide explains what to prepare first.

Scene purpose: Introduce the approved product photo as the visual anchor of a product ad.
Subject: The exact product shown in the supplied reference image.
Action: Subtle camera movement around the product; no change to the product itself.
Setting: Extend only the approved background style from the reference image.
Composition: Vertical 9:16 crop that keeps the product fully visible and leaves space above it for captions.
Camera: Slow, stable orbit or push-in lasting [3–4] seconds.
Product accuracy: Preserve the exact shape, color, label, proportions, and logo from the reference.
On-screen text space: Plain area in the upper third for verified text added later.
Avoid: Replacing packaging, generating hands that cover the label, adding unapproved props, or animating product claims.

Generated motion may still introduce product errors. Review the result frame by frame, especially around labels, fine edges, reflections, and accessories.

Example 4: hook prompt with a visible problem

A hook should make a relevant situation recognizable without overstating the product outcome.

Scene purpose: Open on the everyday situation that [product] addresses.
Subject: A simple, non-sensitive task related to [approved use case].
Action: Show the task becoming inconvenient or incomplete in one visible way.
Setting: [Specific real-world setting] appropriate to the audience.
Composition: Vertical 9:16, clear visual focus, no clutter.
Camera: Medium shot followed by a brief close-up of the task.
Product accuracy: Do not show the product yet, or show only the approved package at the end of the scene.
On-screen text space: Bottom third clear for a question such as “[viewer question]?”
Avoid: Dramatic harm, false urgency, exaggerated emotion, or a result that the product cannot support.

The next scene should answer the question with a product use or a factual detail. Do not let a problem-led opening imply that the product guarantees a universal solution.

Example 5: CTA end-card planning prompt

An end card is often better assembled from approved brand assets. If you need a generated transition scene before it, keep the prompt focused on composition rather than text rendering.

Scene purpose: Create a simple final product hero scene before the approved end card.
Subject: Approved [product] on a clean surface with supplied logo asset available for editing.
Action: Product remains still; only a slow camera push-in.
Setting: Minimal background using approved brand colors.
Composition: Product in lower center; leave the upper half empty for a CTA added in post-production.
Camera: Slow and stable push-in.
Product accuracy: Preserve product packaging and proportions exactly.
On-screen text space: Large blank upper area for the approved CTA and legal text, if required.
Avoid: Generating logos, prices, offer terms, badges, or claims as image text.

Add the exact CTA, destination cue, logo, and offer terms in a controlled design tool where they can be checked for accuracy and readability.

Add constraints that make review easier

Useful constraints are specific enough for a reviewer to check. They are not a way to guarantee that a generation system will follow every instruction. Put the most important constraints near the beginning of the prompt and repeat only the details that are essential to product accuracy.

For a packaged product, identify the approved reference image, the label orientation, the product color, and the allowed props. For a software interface, identify the exact screen state that is available, what data must remain fictional or hidden, and whether a later editor will add interface text. For a person, identify the role needed in the scene without creating a false identity, customer history, qualification, or endorsement.

You can attach a compact review note to every prompt:

Required source assets: [file or approved asset description]
Must remain accurate: [three product or brand details]
May vary: [lighting, background treatment, camera distance]
Must not imply: [testimonial, outcome, certification, comparison]
Reviewer: [product / brand / legal-policy owner]

This note is particularly useful when several people are creating variants. It separates a visual preference—such as a warmer background—from a fact that cannot change, such as the product label or an offer term. If an output conflicts with a required detail, reject or revise it instead of trying to correct the mistake through copy alone.

Write prompts for the edit you actually need

Many short ads are assembled from several simple shots. Plan each scene around the edit point. A hook can end on the product entering frame; the demonstration can begin with the same product position; the final product hero can leave the same side of the frame empty for a CTA. These small continuity decisions make the sequence easier to assemble than one long request for a complete advertisement.

For every scene, note the intended duration, first frame, last frame, and whether audio, captions, or a transition will be added later. Keep voiceover and legal text outside the generation instruction unless the workflow has a reliable, approved way to control them. The prompt should describe what the viewer needs to see; the production plan should control what the viewer needs to read.

Prompt review checklist

Before generating a scene, ask:

• Does this prompt have one scene purpose and one visible action?
• Does it name the product details that must not change?
• Does it leave text, logos, prices, and legal copy for a controlled editing step?
• Does it avoid invented testimonials, credentials, reviews, or outcomes?
• Is the product reference asset approved and appropriate for the placement?
• Can a reviewer compare the output against the creative brief?

After generation, review the product, people, setting, captions, spoken words, claims, and destination together. A generated scene is a draft. It may be visually appealing while still being inaccurate, unclear, or unsuitable for the campaign.

Turn prompt examples into a production packet

For each scene, keep the brief, prompt, approved source assets, and review notes together. Label versions by the variable you changed—for example, hook-question-v1 or product-closeup-v2—rather than by vague labels such as final-final.

This process makes it easier to compare alternatives while keeping product facts stable. Start with a concise AI video ad planning brief, create scene instructions, then use UGC Labo to keep the strategy visible before any finished-video production begins.