How to Make Product Videos With AI From One Photo (2026)

An AI product video ad starts from a real photo of the real product. You clean up the photo, place it in a scene, and an image-to-video model adds a few seconds of motion: a slow turn, a pour, a lid lifting, light sliding across the label. Then you add a hook, captions and an end card in your editor.

That order matters. Video models are good at motion and light, but they also redraw whatever they animate. Text on packaging can warp, and a model can make a bottle look bigger or a fabric shinier than it is. An ad has to show the product as it really is, so the photo of the real item stays the source of truth and the AI only adds the movement.

Everything in this guide runs in Spicy Studio: the Background Remover and product templates for the still, and the Video Lab for the motion. It is a lesson in the free Spicy Academy "Make money with AI" track, whether the ads are for your own shop or for clients.

Shoot the product on a plain surface by a window, with the phone at the product's height. Take a few angles, including the front of the packaging straight on.

Cut it out. Run the best shot through the Background Remover. An image costs 4 credits and gives you a transparent cut-out you can reuse in every scene.

Place it in a scene. Either pick one of the product templates (18 credits per image) or describe the scene yourself in the Image Lab with Nano Banana 2.1 and the cut-out attached as a reference image (12 credits at 2K). Ask for the product to stay exactly as it is: "Place the product from the reference image on a light oak kitchen counter, morning window light from the left, shallow depth of field. Keep the product, its shape, colors and label exactly the same."

Check the still before you animate. Compare it with the real item: color, proportions, label text, cap, what's included. A video model will faithfully animate a mistake, so fix it here.

Spicy Studio has three image-to-video models that suit product ads. Pick by the shot, not by the brand name.

Kling 3.0: controlled, repeatable product shots. Takes a start frame and an optional end frame, which gives you the most control over a transformation, such as a closed box at the start and an open box at the end. Clips are 5 or 10 seconds in 16:9, 9:16 or 1:1. Standard (720p) costs 60 credits for 5 seconds and Pro (1080p) costs 80; sound adds 50%. Open Kling 3.0.

Seedance 2.5: longer clips, sound and references. Clips run from 4 to 30 seconds at 480p, 720p or 1080p, with audio generated in the same pass and up to 10 reference images. It also offers 3:4, 4:3 and 21:9 as well as the usual ratios. It is billed per second: 10 credits at 480p, 22 at 720p and 39 at 1080p, so 5 seconds at 720p costs 110. Open Seedance 2.5.

Veo 3.1: cinematic 8-second shots. A fixed 8-second clip in 16:9 or 9:16 with an optional end frame, for hero shots with camera movement and depth. It costs 80 credits at 720p, 110 at 1080p and 200 at 4K. Open Veo 3.1.

A practical routine is to draft in Kling 3.0 Standard or Seedance 2.5 at 480p, keep the take that works, and only then render the final at a higher resolution.

Keep each prompt to one simple sequence of actions, describe the camera, and end by asking for the product to stay unchanged. These work as starting points; replace the product and setting with yours.

Turntable. "The bottle slowly rotates 360 degrees on a white marble surface. Soft studio light. The camera stays still. The product and label stay exactly the same."

Push-in. "Slow camera push-in toward the watch on a dark wooden desk, warm side light, shallow depth of field."

Pour. "Cold brew pours into the glass from the top of the frame, ice cubes move, condensation on the glass, morning window light."

Lid reveal. "A hand lifts the lid of the gift box to reveal the candle inside, then pulls back out of frame. The camera remains still."

Jewellery sparkle. "A soft light sweeps slowly across the ring on a dark velvet stand, catching the facets of the stone. The camera remains still. The ring stays exactly the same."

Float. "The sneaker floats just above a pastel backdrop and slowly turns, gentle shadow below."

In use. "A woman's hands open the notebook, write a short line and close it again. Natural daylight."

Unboxing POV. "First-person view: hands open the shipping box, lift out the headphones and hold them toward the camera."

Texture macro. "Macro shot, the camera glides slowly across the knit texture of the sweater."

Seasonal scene. "Snow falls softly outside the window behind the lit candle, the flame flickers gently, cozy evening light."

Ingredient drop. "Fresh mint leaves and lime slices drop gently around the can, small water droplets on the can."

End-frame transformation (Kling 3.0). Attach the folded jacket as the start frame and the same jacket on a hanger as the end frame, then prompt: "The jacket unfolds and lifts onto the hanger in one smooth movement. The camera remains still."

When text on the label warps, keep the camera still, shorten the clip, or use the clean still as the last frame and put the real packshot on your end card.

The generated clip is raw footage. The ad happens in the edit.

Hook in the first second. Start on motion, not on a logo. Put the main benefit in a short on-screen line.

Captions and one message. Many people watch without sound. One benefit per ad is easier to remember than five.

End card. Finish on the real product photo, the name, the offer and where to buy it.

Sound. Use music you have a license for. Seedance 2.5 and Kling 3.0 can also generate ambient sound with the clip.

Marketplace videos. Product pages and feeds often autoplay video without sound, so let the motion carry the message. If a product video for Amazon or anywhere else shows a photorealistic AI-generated person, follow the disclosure rules further down.

Formats. TikTok, Reels and Shorts are vertical 9:16. Feed placements often use 1:1 or 4:5, and YouTube in-stream is 16:9. Generate in the ratio you need rather than cropping later, and run the finished clip through Video Upscale (16 credits to 1080p) if the platform asks for more resolution than you generated.

These examples use the current credit prices; check the tool before you generate, because prices can change.

One product, three motion ideas. A cut-out (4 credits) and one scene in Nano Banana 2.1 at 2K (12), then three 5-second Kling 3.0 Standard clips (3 × 60). That's 196 credits, about a third of the 600 credits in the €10 Basic plan.

A 15-second Seedance 2.5 ad with sound. 15 seconds at 720p is 330 credits; drafting it at 480p first costs 150.

One cinematic hero shot. Veo 3.1 at 1080p is 110 credits for 8 seconds.

New accounts get 25 free credits, enough for the cut-out and a scene. Plans and one-time top-ups (from €2 for 90 credits) are on the pricing page.

AI makes it easy to produce an ad that looks better than the product. That is where the risk is.

Show the real product. Size, color, materials, what's in the box and what it does must match what the buyer receives. In the US, the FTC requires ads to be truthful and claims to be backed by evidence. Canada's Competition Act and Australia's Consumer Law also prohibit misleading advertising.

No fake results. Don't animate a before-and-after the product can't deliver, and don't show results such as skin, weight or speed changes that you can't prove.

Label AI where the platform asks. TikTok requires ads that are AI-generated or significantly edited to be disclosed (simple lighting, color and background-removal edits are exempt). If the ad shows a photorealistic AI-generated person, New York has required a disclosure since June 2026, and Amazon requires a synthetic-performer tag in listing images and videos. AI UGC Ads covers TikTok and Meta ad labels in detail.

Use rights you have. Use your own product photos or the client's, licensed music and no other brands' logos or characters.

Small shops often have good photos and no video. A short package, such as three 5-second motion clips in 9:16 and 1:1 plus one 15-second edited ad, is easy to scope and quick to deliver. AI Services to Sell covers pricing, contracts and where to find clients, and Amazon Image Requirements & AI Product Photos covers the still images that come first.

This guide teaches a skill. It is not a promise of income or a business opportunity, and it isn't legal or financial advice. What you earn, if anything, depends on your skills, effort, market and costs, and some projects earn nothing. If you advertise or sell services, consumer-protection rules apply in each market you sell to: the FTC in the US, the Competition Bureau in Canada and the ACCC in Australia all treat misleading claims as illegal. None of these regulators has reviewed or endorsed this guide.