How to turn product photos into AI video ad creatives

Image-to-video AI can turn a product photo into a moving shot for an ad. Start with a clear image, ask for one controlled movement, then add your headline, offer, and CTA in a video editor. Check the product throughout the clip before using it.

A slow push toward a travel mug draws the eye closer to the product. A generated hand opening its lid asks the model to invent how the product works. That second shot needs much closer scrutiny.

This guide covers the first job: turning approved photography into usable motion for a product ad. The tool steps follow current Runway and Adobe documentation; the prompts and travel-mug example are suggested starting points, not tested campaign results.

There are two ways to make a still photo move. A video editor can pan or zoom across the existing image. An image-to-video model generates new frames, which can introduce movement within the scene.

Use a normal editor when all you need is a gentle zoom behind a headline. It gives you predictable control over the original photograph. Use generative video when you want environmental movement, a changing camera perspective, or another effect that requires new visual information.

For the last option, see our comparison of AI video generators for Facebook ads. Here, the starting asset is one photograph and the output is a shot you will edit into an ad.

Choose the original product photo where possible. A compressed screenshot of an existing ad brings small text, interface clutter, and compression artifacts into the generation.

Keep the headline, discount badge, and CTA out of the source image. Add them later as editable layers. Product packaging may already contain text; inspect that text carefully in the output instead of assuming the first frame guarantees it will stay correct.

Before uploading, check that:

• The product matches the version you sell, including its lid, handle, color, and accessories.
• The important surfaces are sharp enough to inspect.
• The framing leaves room for the intended motion and the copy you will add.
• You have an approved source for every visible claim and permission to use the photo.

Decide the intended placement before composing the shot. Use the Meta ad sizes guide to plan the canvas, then check the crop offered by the generation tool. A product close to an edge gives the model very little room for a push-in.

Save the original separately. You will need it when a later frame looks plausible but something about the product feels wrong.

Both tools accept an image as a starting frame. Choose based on the controls you need and the editing tools your team already uses. The steps below follow their documented workflows.

Runway: direct the movement with a prompt

Runway's Gen-4.5 documentation describes image-to-video generation with a required image and text prompt.

1. 1.Open Gen-4.5 from Apps View, or select it in Video Tool Mode.
2. 2.Upload the prepared product image.
3. 3.Write a prompt describing the movement you want.
4. 4.Choose the available duration and aspect ratio that fit the shot, then generate.
5. 5.Review the result before downloading it for editing.

Runway documents 2–10-second durations for this model. Start with a short, simple shot so you can identify which instruction caused a problem.

Adobe Firefly: use a starting frame and camera controls

In Firefly, open Generate video and select the Firefly Video model. Under Frame, upload your image as the first frame. Add a motion prompt, review the available settings, and generate the clip.

Adobe's image-to-video instructions document camera-motion controls when only a first frame is supplied, including zoom and static options. You can also supply a last frame, but doing so disables the separate Motion controls.

For a first product shot, use one starting frame. Add a last frame when you have a specific approved ending composition to work toward, and inspect the transition between them.

Runway's image-to-video prompting guide recommends focusing on movement because the image already provides the subject, composition, lighting, and style. Begin simply and refine the instruction after reviewing the output.

For a travel mug photographed on a tabletop, try this original prompt:

The camera slowly pushes toward the mug in one continuous shot. The mug stays upright and still on the tabletop. The movement eases to a stop with the full lid and handle visible.

If the scene includes a curtain, an alternative is:

The camera stays fixed. The curtain in the background moves gently in a light breeze while the mug remains still. Movement is subtle and continuous.

Keep the request this narrow until you have a usable shot. Add the sales message during editing, where you can control every word.

Change one instruction between attempts. If the push-in cuts off the handle, reduce the movement or give the source image more room. If the whole mug changes shape, try a fixed camera and background movement. A more elaborate prompt is not always the repair you need.

Play the clip at normal speed, then scrub through it beside the original photo. Pay particular attention to the middle of a camera move. An accurate opening frame does not tell you what happens afterward.

Use a short review sheet:

Do not use generated product behavior as proof. If the ad needs to demonstrate a leak-resistant lid, use real demonstration footage supported by the product facts. A plausible synthetic pour tells the buyer nothing reliable about that lid.

If a shot keeps failing, animate the approved still in an editor. You can also place the original product cutout over a separate moving background, though lighting and edges will need attention. Set a small retry budget before you start. One five-second shot can consume the afternoon.

Import the accepted shot into your editor. Keep the logo, headline, offer, and CTA separate from the generated footage so a copy change does not require another generation.

Here is an illustrative eight-second edit for a travel mug. The timing is an editing example, not a performance benchmark, and the copy assumes the product fits the description.

The final still holds the offer steady while the viewer reads it.

If you add narration, caption the spoken message. Check the edit with sound off and make sure someone can still understand what is being sold. Also read the overlay text at phone size; a line that looks comfortable on a large editing monitor can be too small in the placement.

Export for the intended placement and review the assembled ad using the Meta ad preview workflow. Include the real identity, destination, and CTA in that check, since the uploaded video is only part of what people will see.

Write down what the motion is supposed to improve. For example: "A slow product push-in will make the opening easier to notice without hiding the handle."

Keep the offer, landing page, and main message consistent if motion is the variable you want to study. Changing all of them creates a different ad, but makes it harder to understand the result.

A moving version can earn more attention and still fail to produce useful visits or sales. Read the outcome against the campaign objective, and use attention metrics to diagnose the opening rather than declare a winner. Our guide to static ads versus video ads covers when each format has a useful job.

Adrio helps teams research competitor ad patterns and create editable static ad variations. Use that stage to settle the message, product presentation, and visual direction before handing an approved image to a video tool.

You can prepare that image in Adrio, then use Runway or Firefly for motion and an editor for the final ad. Keep the original image and approved copy with the exported clip so the next revision starts with the same product facts.

Can I turn a finished static ad into an AI video?

You can use it as an input, but a flattened design includes text and logos that may change in generated frames. Prefer a clean product image, then rebuild the headline and offer as editable overlays.

How do I keep the product exactly the same?

A prompt cannot guarantee exact preservation. Limit the motion, compare the full clip with the original, and reject altered products. If exact pixels matter, animate the original image or composite it in an editor.

Do I need a separate video editor?

You need an editing step for the message, timing, captions, and final output. That may happen inside a tool you already use. Generating a shot and finishing an ad are separate tasks even when one application handles both.