AI video ad guidesMakeup ads

How to make b-roll for ads with AI

The easiest way to make AI b-roll is to animate a real product photo, one shot at a time. In Adobe Firefly, choose Generate Video, upload the photo as the first frame, and prompt a single slow movement. Keep real footage for anything that has to prove how the product works.

Updated November 20, 20266 steps5 prompts3,632 real ads

Make my video ads โ†’

$1 today. Cancel anytime before day 7.

The steps
  1. 1Write a shot list firstPick 3 to 5 visuals that support your voiceover: a product reveal, a detail close-up, a lifestyle shot, and a final product shot with room for a call to action.
  2. 2Upload a clean product photoIn Firefly, choose Generate Video and add your photo as the first frame. Use a sharp, evenly lit image with the label facing the camera.
  3. 3Prompt one movement per shotAsk for a slow push-in, a gentle sideways slide, or light moving across the product. Say that the shape, color, and label stay unchanged.
  4. 4Generate several takesRun each shot a few times and keep the cleanest one. Throw away any take where the label or color shifts.
  5. 5Cut 1 to 3 seconds per clipPlace the best moment of each clip over your voiceover in a 9:16 timeline. Add your logo, offer, and CTA in the editor, not in the prompt.
  6. 6Check details, then publishWatch every clip at full size for warped text, odd hands, and color drift. If a product keeps changing, use the original photo with a slow zoom instead.
On this page
  1. The steps
  2. Real makeup video ads
  3. What AI b-roll should and should not do in an ad
  4. Build a shot list before you open any tool
  5. Animate a product photo in Adobe Firefly
  6. Use text-to-video only for generic cutaways
  7. What long-running makeup ads teach about b-roll
  8. Edit in 9:16 and check every frame
  9. Prompts to copy
  10. Which method to use for each b-roll shot
  11. Mistakes to avoid
  12. Questions

Real makeup video ads to model yours on

3,632 video ads from 365 advertisers in the Meta Ad Library, read October 5, 2026. The median one ran 22 days, and an ad still paid for after months is the closest public sign that it works. Live ads first.

  • Ran under a day
  • Ran 7 days
  • Ran 2 days
  • Ran 7 days
  • Ran under a day
  • Ran 4 days
  • Ran under a day
  • +3,620 more makeup video ads

    Hooks, landing pages and run length for each

  • Ran under a day
  • Ran 6 days
  • Ran 6 days
  • Ran 4 days

Every creative, with its copy, format and run length

The openers that ran longest

  • 40% OFF + FREE Tint โ€“ Lashes Youโ€™ll LOVE!Luxe Cosmetics ยท ran 504 days
  • Chat With UsIsabella Garcia International ยท ran 455 days
  • Realize Sonhos, Seja ConsultoraNatura ยท running 434 days
  • Brushes Are Back in StockMorphe ยท ran 429 days

What AI b-roll should and should not do in an ad

B-roll is the footage that plays over your voiceover or captions while the main message lands. In a short video ad it carries most of the visual weight: the product reveal, the texture close-up, the mood of the final shot. These shots are brief, silent, and loosely defined, which is exactly where AI video performs best.

AI is weaker wherever the product has to be exact. Small label text, fine hand movement, application steps, and before and after results are all claims. A viewer who spots a warped label or a shade that shifts mid-clip stops trusting the whole ad. Keep real footage for anything that proves how the product works, and use AI for the shots around it.

A simple rule keeps you safe. If your actual product is visible, animate your own photo with image-to-video so the item stays true to life. If the shot is generic, like soft window light, a vanity table, drifting steam, or fabric moving, use text-to-video. The first protects accuracy, the second saves you a shoot.

Build a shot list before you open any tool

Most weak AI b-roll comes from generating clips with no plan. Start from the script instead. Read the voiceover line by line and write one visual per line. A 15 to 20 second ad usually needs 4 or 5 shots, each on screen for 1 to 3 seconds. That pace keeps the edit moving and hides small AI flaws.

Give each shot a job. Shot one is the hook, a product reveal that matches the opening line. Shot two is a detail close-up of texture, finish, or material. Shot three is a lifestyle or atmosphere cutaway. Shot four is the final product shot with empty space for your headline and button. If you sell makeup, that could be a compact lid catching light, a bullet of lipstick, a vanity scene, and a packshot.

Write the movement next to each shot before you generate anything. Examples: slow push-in, slow slide left, light sweep, static with background drift. One movement per shot is the habit that keeps products stable. Also note which shots need your real item. Those become image-to-video from a photo, and the generic ones become text-to-video.

  • One visual per voiceover line
  • 4 to 5 shots for a 15 to 20 second ad
  • One camera movement per shot
  • Mark each shot as photo-based or generic

Want makeup video ads like these for your product?

The makeup video ads that kept running, read from the ad libraries. Every change waits for your approval.

Animate a product photo in Adobe Firefly

Open Firefly and choose Generate Video. Upload your product photo and set it as the first frame, so the clip starts from your real item instead of a lookalike. Choose a vertical 9:16 aspect ratio if the option is offered for your model, because cropping a horizontal clip later wastes pixels and cuts off your product.

Your photo matters more than your prompt. Use a sharp image on a plain or softly lit background, with the label facing the camera and nothing cluttered near the edges. Leave empty space above or below the product if you plan to add a headline. A busy photo gives the model more reasons to invent details and drift.

Keep the prompt short and physical. Name the camera move, the light, and what must not change. The copy-paste prompt from the prompt section does this in five short sentences. Generate three or four takes per shot. Compare them at full size and pick the one where the label stays crisp from the first frame to the last. Save the winner and move on.

Use text-to-video only for generic cutaways

Text-to-video works for shots where no specific product appears. Think of a bathroom shelf in morning light, powder floating in sunlight, a hand-free flat lay of brushes, or fabric rippling. Describe the scene, the light, the lens feel, and the camera move. Add that there should be no text, no logos, and no people if you want a clean cutaway.

Use these shots sparingly. They set the mood and give your edit rhythm, but they do not sell the product. A good ratio is one generic cutaway for every two or three product shots. More than that and the ad feels like stock footage with a product stuck on the end.

Match the look of your cutaways to your photo-based shots. Use the same time of day, color temperature, and background tones in every prompt. Mixing a warm golden scene with a cold blue packshot makes the edit feel stitched together, even when each clip looks good alone.

What long-running makeup ads teach about b-roll

The Sprites ads library shows 3,632 makeup video ads from 361 advertisers in the Meta Ad Library, with a median run of 21 days. The ads that ran far longer lead with a clear promise, then let the visuals back it up. Rise Luxus opened with "Buy 1 Get 1 FREE Today!" and ran 372 days. Sheryl Davis opened with "Flawless Skin In Seconds" and ran 294 days.

Prime Prometics ran "Foundation You're Sought, Made Real." for 292 days, and Wonderskin ran "The Blush That Becomes Uniquely Yours" for 185 days. The pattern is a single plain line that names an offer or a result. Nothing in those openers needs a complicated visual. That is good news for AI b-roll, because the footage only has to support one idea.

So build each shot to prove the line. An offer opener wants a clean product reveal and a final packshot with space for the deal. A result opener like "Flawless Skin In Seconds" needs real footage for the result itself, with AI shots covering the product and the mood. Pair the opener with the right shot, then cut quickly. For more patterns, see makeup video ads.

Edit in 9:16 and check every frame

Drop your clips onto a vertical timeline over the voiceover. Use only the best 1 to 3 seconds of each clip, usually the smooth middle where the move is steady. Trim the first and last half second, where drift and artifacts are most likely. Cut on a beat of the voiceover so each visual lands with its line. Check safe zones in the Meta ad specs so captions are not hidden by interface buttons.

Add your logo, offer, price, and CTA in the editor. Never ask the AI to render text. Generated lettering is often misspelled or warped, and the editor gives you crisp type you can change in seconds. Place the headline in the empty space you left in the photo, and keep the button copy short, like Shop now.

Then review every clip at full size before export. Look for warped or shifting labels, color changes on the product, melting edges, extra fingers, and background objects that appear or vanish. If a shot fails three or four takes in a row, stop fighting it. Use the original photo with a slow zoom in your editor. A stable still beats a moving mistake.

Prompts to copy

Paste each one as written, then change the product, the setting and the light to match your own footage.

  1. Hook: product reveal (image-to-video)

    Vertical 9:16 product close-up. Slow camera push-in. Soft morning light moves across the background. The product stays still, with its shape, color, and label unchanged. Leave open space above the product for a headline. No added text, no people, no logos added.
  2. Detail close-up: texture and finish

    Vertical 9:16 macro shot of the product from the uploaded photo. Camera glides slowly sideways, a soft highlight sweeps across the surface. Shallow depth of field, clean neutral background. Packaging shape, color, and label stay exactly the same. No added text, no hands, no extra objects.
  3. Lifestyle cutaway (text-to-video)

    Vertical 9:16 shot of a bright bathroom vanity in warm morning light, empty glass shelf with a small plant and soft towels. Slow push-in, gentle dust particles drifting in the sunbeam. Calm, premium mood, shallow depth of field. No people, no text, no logos.
  4. Atmosphere: light sweep on the product

    Vertical 9:16 static camera on the product from the uploaded photo, centered on a soft pastel surface. A slow band of window light travels left to right across the scene. The product does not move or change. Keep the top third empty for a headline. No text, no extra items.
  5. Closer: packshot with CTA space

    Vertical 9:16 final product shot from the uploaded photo. Very slow pull-back, product stays centered in the lower two thirds of the frame. Clean soft background with subtle gradient light. Leave the top third empty for an offer and button. Label, shape, and color remain unchanged. No added text.

Which method to use for each b-roll shot

ShotBest methodTime in final cutMain risk
Product revealImage-to-video from your photo1 to 2 secondsLabel warping
Texture close-upImage-to-video, slow slide1 to 2 secondsColor drift
Mood cutawayText-to-video1 to 3 secondsLook mismatch with product shots
Final packshotImage-to-video or photo with slow zoom2 to 3 secondsProduct changing shape
Application or resultReal footageAs neededOverpromising if faked

Mistakes to avoid

  • Generating clips before writing a shot list, so nothing matches the voiceover.
  • Asking for several camera moves in one prompt, which makes the product warp.
  • Uploading a cluttered or low-resolution product photo as the first frame.
  • Using text-to-video for shots that need to show your exact product.
  • Asking the AI to render your logo, offer, or CTA instead of adding text in the editor.
  • Publishing without watching each clip at full size for label and color changes.

Rather have the videos made for you?

Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. You never write a prompt.

Apply to Director by Sprites โ†’

Questions

What is AI b-roll for ads?

It is short supporting footage generated by AI, such as product close-ups, camera moves, and atmosphere shots, that plays over your voiceover or captions. You can make it from a product photo or from a text prompt. It supports the message and does not replace proof footage.

Can I make b-roll from a product photo?

Yes. In Adobe Firefly, choose Generate Video, upload your photo as the first frame, and prompt one slow movement. This keeps the clip anchored to your real product. Generate a few takes and keep the cleanest.

How long should each b-roll clip be in the ad?

Use the best 1 to 3 seconds of each clip. Short cuts keep the pace up and hide small AI flaws. The final packshot can run a little longer so the CTA is readable.

What aspect ratio should AI b-roll be for ads?

Use 9:16 for Reels, TikTok, and Shorts. Generate vertically when you can instead of cropping later. Check the Meta ad specs for safe zones before you place text.

Should I use AI b-roll or real footage?

Use both. AI is good for close-ups, movement, and mood. Real footage is better for anything that shows how the product works or what results look like. If AI keeps changing your product, use the original photo with a slow zoom.

Can someone else make the b-roll for me?

Yes. Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. If you want to do it yourself, start with the video ads guides.

Keep going

Get makeup video ads made for your brand

  • The makeup video ads that kept running, read from the ad libraries
  • Finished, on-brand video ads, reviewed by a human
  • Every change waits for your approval