Guides

AI UGC Ad Prompts: Shot-by-Shot Scripts for Sora, Veo and Kling

By the Prompt Shelf team · Updated November 17, 2026 · 9 min read

If you run a Shopify or DTC brand, or you freelance making ads for one, you have probably stared at an empty prompt box and hoped Sora or Veo would guess what a good product ad looks like. AI UGC ad prompts are fill-in-the-blank instructions that tell an AI video tool exactly what to shoot: the subject, the action, the camera, the lighting and the style, in the order a short vertical ad needs them. This guide shows you how to plan the ad first, write the script, convert it into prompts, and test the variants.

Quick answer: AI UGC ad prompts are fill-in-the-blank instructions that tell an AI video tool what to shoot: product, action, camera, lighting and style, for a 15-30 second vertical ad. Plan the angle in ChatGPT, Claude or Gemini, write the script, convert each shot into a prompt, generate a few versions, and keep the clips where the product looks accurate.

What AI UGC ad prompts are and why they matter

An AI UGC ad prompt is a short, structured instruction that tells a text-to-video or image-to-video tool exactly what to film, so the result looks like a creator-made product ad instead of a generic clip.

UGC means user-generated content: the loose, phone-shot style that performs on TikTok, Instagram Reels and YouTube Shorts. The prompt is how you get that style on demand. Instead of describing a vibe, you name the subject, the action, the setting, the camera, the lighting, the style, the duration and the things to avoid. Each element is one line a tool can actually read.

The structure matters for two reasons. Testing hooks and angles stops depending on booking a creator for every idea. And the same prompt shape works in Sora, Veo, Kling and Runway, so you can switch tools without rebuilding your process.

Step by step: from product to tested ad

  1. Write the product-to-angle brief. Open ChatGPT, Claude or Gemini and give it the product, the price and offer, the target buyer, the problem in the buyer's own words, and the proof you can actually show. Ask for five distinct ad angles in a table.
  2. Pick two angles and draft hooks. Choose the angles with the clearest proof, then ask for ten hooks for one angle: two visual, two question, two pattern interrupt, two demonstration and two myth-busting.
  3. Write the 15 or 30 second script. Turn the best hook into a script with a shot list. Fifteen seconds usually holds three shots; thirty seconds holds five or six.
  4. Run the claims and disclosure check. Read every line and mark anything that states a result, a number or a guarantee you cannot support. Rewrite it as a brand or creator statement, never a customer testimony.
  5. Convert each shot into a prompt. One shot, one prompt, in the order subject, action, setting, camera, lighting, style, duration, avoid.
  6. Generate 2-4 versions per clip. Keep the versions where the product looks accurate, the hands are normal and nothing invents a logo or a feature.
  7. Edit the clips together. Assemble in CapCut, Premiere or your usual editor. Add your real price, code and logo as text, and label the ad as your platform requires.
  8. Test variants. Run the A/B prompt on your two angles and compare how many viewers are still there at five seconds, not just which clip gets more likes.
Prompt pack for AI video and UGC ad creators
AI UGC Ad Video Prompts for Sora, Veo and Kling: Product Demo and Hook Scripts
Shot-by-shot prompts and 15-30 second scripts for UGC-style product ads you can generate with AI video tools.
$24 · See what's inside →

Copy-paste prompts

When to use: you have a product and a platform but no angle yet. Run this in ChatGPT, Claude or Gemini before you open a video tool.

You are a direct-response creative strategist who plans short UGC-style video ads for DTC brands.

Product: [PRODUCT NAME AND WHAT IT DOES]
Price and offer: [PRICE AND CURRENT OFFER]
Target buyer: [WHO BUYS IT, AGE RANGE, LIFESTYLE]
Main problem it solves: [THE PROBLEM IN THE BUYER'S WORDS]
Proof I can show: [REAL FEATURES, SPECS, GUARANTEE, REAL REVIEW COUNT]
Platform: [TIKTOK / INSTAGRAM REELS / YOUTUBE SHORTS / META FEED]

Task: propose 5 distinct ad angles for a 15-30 second vertical video. Distinct means a different reason to buy, not the same idea reworded.

Output a table with columns: Angle name | Core promise | Pain it targets | What the video must show | Proof needed | Overclaim risk (what NOT to say).
Then recommend the 2 angles to test first and explain why in 3 sentences.

Use only the facts I gave you. If a key fact is missing, ask me before you write the table. Do not invent statistics, reviews or certifications.

When to use: you picked an angle and need the first three seconds. Two hooks of each type, written so the camera description can go straight into a video tool.

You are a UGC ad scriptwriter who specializes in the first 3 seconds of short vertical videos.

Product: [PRODUCT NAME]
Chosen angle: [ANGLE, E.G. "SAVES 10 MINUTES EVERY MORNING"]
Target buyer: [WHO]
Tone: [CASUAL / FUNNY / CALM / ENERGETIC]

Write 10 hooks, two of each type: visual, question, pattern interrupt, demonstration, and mistake or myth-busting.

For each hook give: on-screen text (7 words max) | spoken line (under 12 words) | what the camera sees in the first 2 seconds, written so I can paste it into an AI video tool.

Mark with [CHECK] any hook that makes a claim I would need to prove. Do not write hooks that pretend the speaker is a verified customer. If a product fact is missing, ask me before you write the hooks.

When to use: you need a hook shot that shows the product without a person. Fill in the brackets, paste into Sora, Veo or Kling, and generate at least three versions.

Vertical 9:16 video, 4 seconds.
Subject: [PRODUCT, WITH COLOR, MATERIAL AND SIZE].
Action: the product drops from above into frame and lands on [SURFACE, E.G. A MARBLE COUNTERTOP], settles with a small bounce, then the camera slowly pushes in.
Setting: [SETTING, E.G. BRIGHT MODERN KITCHEN, BLURRED BACKGROUND].
Camera: macro close-up, 50mm look, locked off for the drop, then slow push-in; shallow depth of field.
Lighting: [LIGHTING, E.G. SOFT MORNING WINDOW LIGHT FROM THE LEFT].
Style: crisp, realistic product commercial shot on a phone-quality-plus camera, natural colors.
Sound: a light, satisfying landing sound, no music.
Leave the top third of the frame clear for on-screen text.
Avoid: warped or melting product shape, extra products, invented text or logos on the product, hands, people.

If any bracketed detail is missing, tell me what you need and I will fill it in before you generate.

Worked example: before and after

Before, a prompt that tells the tool almost nothing:

Make a cool ad for my insulated water bottle. Show someone at the gym. Make it feel authentic and get lots of views.

That asks for an outcome, not a shot. The tool chooses the angle, the lighting, the framing and the claims, and you get a clip you cannot art-direct.

After, the same idea as a structured prompt:

Vertical 9:16 video, 5 seconds, handheld phone-camera feel.
Subject: a woman in her late 20s in a grey tank top, standing by gym lockers, holding a matte sage-green insulated bottle.
Action: she twists the lid off, takes a drink, then smiles at the camera.
Setting: bright gym corridor, blurred background, a few bags on a bench.
Camera: medium close-up, slight natural handheld shake, eye level, 26mm smartphone lens look.
Lighting: bright overhead gym light, natural and ungraded.
Style: authentic UGC, realistic skin texture, no cinematic grading.
Spoken line: "I look forward to water now."
Avoid: exaggerated expressions, readable brand names, extra fingers, product claims on screen.

The second prompt gives you a shot you can control. Change one bracket and you have a variant.

Shot plan at a glance

Shot Job it does Typical length Prompt focus
Hook Stops the scroll in the first 3 seconds 3-5 s Action and setting, clear top space for text
Problem Makes the viewer feel the pain 3-5 s One clear moment, handheld feel, no product
Demo Shows the product doing the thing 5-8 s One step only, slow deliberate movement
Detail Proves the quality 3-5 s Macro glide along one feature
Talking Delivers the reasons to buy 6-8 s Presenter, spoken line under 20 words
End card Gives the offer and the call to action 3-5 s Turntable or mood shot, room for price and code

Common mistakes

  • Prompting the whole ad in one request. Fix: write one prompt per shot and keep each under about 120 words.
  • Letting the tool invent product details. Fix: use image-to-video with your own product photo, and name the materials, colors and features in the prompt.
  • Writing claims you cannot support. Fix: run the claims check and mark anything unproven as [CHECK] before you generate.
  • Making the AI presenter sound like a real customer. Fix: have the presenter speak as the brand or a creator, and follow your platform's AI-content labeling rules.
  • Shipping the first generation. Fix: generate 2-4 versions per clip and keep only the ones where the product and hands look right.
  • Skipping the hook. Fix: put the result or the problem on screen in the first second, not the logo.

Frequently asked questions

Which AI video tools do these prompts work with?

They are written for text-to-video and image-to-video tools such as Sora, Veo, Kling and Runway. The structure (subject, action, setting, camera, lighting, style, duration, avoid) carries across tools. Tools differ in clip length, audio and lip-sync, so a tool without built-in speech needs spoken lines added later as voiceover. Check each tool's limits before you plan a shot list.

Will the AI keep my product looking exactly like the real one?

Not always. Text-only prompts can change shapes, colors or labels. For accuracy, use the image-to-video prompt with your own product photo, generate several versions, and film your real product for any close shot where the details matter. Treat the AI output as a first draft you art-direct, not a finished shot.

Can I use AI-generated presenters in ads?

Many brands do. The presenter should speak as the brand or a creator, not pretend to be a real customer, and you should follow your platform's ad and AI-content labeling rules. A prompt pack can include a claims and disclosure check, but that is practical guidance, not legal advice.

What exactly do I get in the prompt pack?

Thirty text prompts in five sections, each with a title, a "when to use it" line, the prompt with fill-in placeholders, and a filled-in example. No video files or software are included. You paste the prompts into the AI tools you already use and edit the results in your usual editor.

How long should each clip be?

Most vertical ad clips run 3-8 seconds, and a 15-30 second ad is usually three to six clips edited together. Short clips are easier to generate well and to re-cut into variants. Keep the hook at 3-5 seconds and give the demo shot the most time.

Next step

If you want the full set, AI UGC Ad Video Prompts for Sora, Veo and Kling: Product Demo and Hook Scripts carries 30 prompts across planning, hook shots, demos, talking shots and end cards, each with a filled-in example. Once you have the ad, Video Repurposing Prompts for Solo Creators helps you cut one long recording into a week of short clips.

Get the full pack

More guides

Free prompts.

Two from every pack, as one PDF. Then one email when a new pack is out.

We send one confirmation link first.