How to build an ad around a single product photo

Last updated 2026-08-25 · 7 min read

What is the short version?

A product photo is not an ad, because the photo has its own background and its own lighting and neither belongs in your film. Separating the product from its picture is the step that turns one flat image into something a camera can move around.

1. Go to Vizard Agent and upload the product photo.
2. Paste the feature list exactly as you have it written.
3. Say the mood — technical, premium, warm, energetic.

What do you need before you start?

One photo and your copy. Vizard Agent handles the cutout, the context footage and the narration voice itself, so nothing visual needs preparing. What it cannot supply is accuracy, so the claims in your feature list have to be right before you paste them in.

• The product photo. One, reasonably sharp.
• The feature list. However it is already written.
• The mood. It sets the music and the voice.
• The audience. Trade, consumer, or technical.
• The length. Thirty seconds suits a feature list.

What do you type into Vizard Agent?

Paste the features and say the mood. Vizard Agent turns a feature list into narration and finds footage that matches each claim, so the useful thing you can add is context about who is watching and how it should feel.

Prompt

Variants worth knowing:

• A transparent cutout. So the product can sit on any background.
• Stock b-roll behind it. Context the photo cannot supply.
• Feature callouts. Timed to the word that names each one.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real ad built from one product photograph. The stretch worth watching is the middle, where it cut the product out, composited it onto the intended background, looked at the result, and re-segmented when the edge was not clean enough.

1. Inspects the attached product image and loads the product-ad playbook.
2. Lists the available generation tools and the effect library.
3. Searches stock footage for the context the photo lacks — the setting, the product in use.
4. Checks music and voiceover options and searches for candidate tracks.
5. Analyses the photo's background and extracts a transparent cutout of the product.
6. Composites the cutout onto the intended background and checks the edge quality by eye.
7. Re-segments the product from the original image and verifies the transparent edge again.
8. Downloads the stock clips and music concurrently, then searches for a narrator voice.
9. Generates the narration, pulls its word timings, and cuts the ad so each feature lands on the word that names it.

Steps 6 and 7 are the ones that decide whether this looks like an ad or like a photograph pasted onto a video. A cutout that seems fine on white falls apart on black, which is why it gets tested on the background it will actually sit on.

What does the result look like?

A short product ad in which the product itself floats free of its original photograph — composited onto a dark background, surrounded by stock footage of the context it belongs in, with narration running through the feature list and each claim landing on screen as it is spoken.

Because the product is a transparent cutout rather than a rectangle, it can move, scale and sit in front of footage. That single change is what separates this from a slideshow with a voice over it.

When does this not work well?

Everything here rests on one photograph, and Vizard Agent can only cut out what that photograph actually contains. These are the cases where the source image sets a hard ceiling on the result, no matter how carefully the rest of the ad is built.

• A soft or low-resolution photo cuts out badly. The edge is where it shows.
• Complex edges are hard. Cables, mesh and transparency fight segmentation.
• One angle stays one angle. A cutout cannot be turned around.
• Busy backgrounds bleed. Colour from the original can cling to the edge.
• Your claims are yours. What the feature list asserts is your responsibility, not the edit's.

How do you fix a result that came back wrong?

Point at the edge or the claim that is wrong. Vizard Agent keeps the cutout, the composite tests, the stock selections and the narration word timings, so the segmentation can be redone or a single feature re-timed without rebuilding the whole ad.

• "The edge looks rough." Re-segmented and re-checked on the same background.
• "That feature is wrong." The narration is regenerated with the corrected copy.
• "The b-roll does not match." Re-selected from the stock searches already run.

How does Vizard Agent compare to doing it yourself?

By hand this means masking the product in an image editor, exporting a transparent file, finding stock that suits it, writing and recording narration, then timing callouts to the words. The masking alone is an hour if the edges are awkward.

Common questions

How many photos does Vizard Agent need? One is enough. Vizard Agent separates the product from its background rather than needing several angles.

What makes a good source photo? Sharp, well-lit, and shot against something the product contrasts with clearly, so Vizard Agent has a clean edge to find.

Will the cutout edge look clean? Vizard Agent tests it on the background it will actually sit on, and re-segments if it does not.

Can the product sit on video rather than a colour? Yes. Vizard Agent composites the transparent cutout over stock footage as easily as over a flat background.

Does Vizard Agent write the narration? Yes, from your feature list, and it pulls word timings so callouts land accurately.

Why re-cut the product out twice? Because a mask that looks perfect against white often shows a pale fringe against black. Checking the composite on the real background and going again is faster than discovering it after the ad is assembled.

Can I supply my own music? Yes. Otherwise Vizard Agent searches the library for tracks that suit the stated mood.

How long should a feature-list ad run? Around thirty seconds. Beyond that the claims start to blur into each other.

Does it check the product name is right? Vizard Agent renders what you give it, so the spelling in your brief is what appears.

Can I get several versions from one photo? Yes. The cutout is reusable, so Vizard Agent can build a vertical cut, a square cut and a wide one from the same segmentation without redoing the masking work each time.

Does Vizard Agent check the finished ad? Yes. It reviews the composite and the callout timing before delivery.