A listing with one good video answers what the photos cannot: how big the thing really is, how the lid opens, how it catches the light when you turn it. You do not need a studio day to get there. An image-to-video model takes a listing photo you already have and moves the camera or the scene for a few seconds, which is often enough for a short product clip.
Most sellers already own the raw material. A listing usually has a main shot on a clean background, a few angles, a scale shot and maybe a lifestyle image. Each of those is a still frame that an image-to-video model can start from. Instead of booking a shoot, you upload the photo, describe a small movement in plain words, and get a clip that opens on exactly the picture shoppers already see.
That starting-frame behavior is the reason this works for products. The models on Picasso IA that accept a photo as the first frame, such as kling-v2.6 and wan-2.6-i2v, keep the visual direction of the picture and add motion on top of it. You are animating your own photography, not asking a model to imagine your product from a description. The catalog has 502 models and 121 of them sit in the text-to-video category, so there is room to compare engines on the same photo; the full catalog lists them.
The cost of trying is low. New accounts get free credits, a clip costs credits that depend on the model and the settings, and you can run the same photo through two engines before deciding. Current plans are on the pricing page, which is the only place to trust for numbers.
Not every photo should become a video. The table below matches the kinds of images a listing already has to the motion that usually suits them, and to the thing most likely to go wrong.
| Source photo | Motion to ask for | What to watch for |
|---|
| Main shot on white background | Slow orbit or gentle push-in | Edges that wobble, shadows that drift |
| Detail close-up (texture, stitching, port) | Slow slide across the detail | Texture turning smooth or plasticky |
| Scale shot with a hand | Light movement of the hand, steady camera | Fingers that change shape or count |
| Lifestyle scene | Soft camera drift, small ambient motion | Background items appearing or vanishing |
| Packaging and contents flat lay | Top-down slow zoom | Items rearranging themselves |
A good habit is to pick the one or two photos that carry the most selling information and animate only those. A clip built from the main shot and the best detail close-up is usually clearer than a clip that tries to move every photo in the set.
This is the shortest honest path from a listing photo to a clip you can review. It assumes you have a sharp, well-lit image of the product and ten minutes.
- Pick the source photo. Choose a sharp, evenly lit image where the whole product is visible. Resolution matters because the model starts from this exact frame, so fix a soft photo first with an upscaler such as clarity-pro-upscaler.
- Open an image-to-video model in the toolkit. In the toolkit, choose a model that takes a start image, such as kling-v2.6 or wan-2.6-i2v, and attach the photo as the first frame.
- Describe one small movement. Write the camera move and nothing else: slow orbit to the left, gentle push-in, soft light shift across the surface. Pick the duration the model offers, for example 5 or 10 seconds on kling-v2.6.
- Generate two or three takes. Same photo, same prompt, a different seed or a different engine. Compare them side by side instead of trusting the first result.
- Review it frame by frame against the photo. Scrub through the clip and compare the product with your listing images: shape, color, buttons, ports, printed details. Keep only a take that matches, and trim or discard the rest.
For more on the general technique, turn photos into videos goes through the basics, and AI product demo videos covers clips that go beyond a single camera move.
The prompt for a product clip is the opposite of a creative brief. You already have the product; the prompt should only describe what moves. Name the camera behavior first (slow orbit, push-in, slide), then the light, then anything ambient. Leave out adjectives about the product itself, because every descriptive word is a chance for the model to redraw something.
Use the negative prompt field when the model offers one, as kling-v2.6 and wan-2.6-i2v do. Typical exclusions for a product clip are extra objects, text overlays, changing colors and distorted shapes. Keep the movement slow, since fast motion is where models start to invent geometry.
Aspect ratio is a decision to make before you generate. On kling-v2.6 the aspect ratio follows the start image when you provide one, so crop the photo to the frame you want first. Square, vertical and widescreen crops can all be prepared from the same original, and the video resizing page covers reformatting a finished clip.
Check the marketplace's current video requirements before you upload. Length, file size, format, aspect ratio and what the video may or may not show are set by the marketplace and change over time, so read the seller help pages for your category instead of trusting a number from a blog.
If the listing photos themselves are not ready, start there: Amazon listing photos covers the stills, and a clip is only as good as the picture it starts from.
Shoppers watch a product video to find out what they are going to receive. A clip that flatters the item beyond the truth gets returns, bad reviews and sometimes a rejected listing. These five mistakes are the common ones:
- Letting the product change: if a button moves, a color shifts or a port disappears between frames, the clip is showing a product you do not sell, so reject that take.
- Adding features that are not there: a model asked for a glowing screen or a pouring liquid will happily draw one, so never prompt for behavior your product does not have.
- Faking scale: a sleek clip that makes a small item look large misleads the buyer, so keep a scale reference or the real dimensions visible somewhere in the listing.
- Moving too fast: quick spins and cuts make models invent surfaces, so prefer a slow move that keeps the item recognizable.
- Skipping the review: a generated clip can look fine at a glance and still contain one wrong detail, so watch it at full size and compare it with the photos before it goes anywhere.
Image-to-video is good at motion around a product. It is weaker at anything that depends on the product behaving correctly. A photo cannot teach a model how your hinge folds, how your zipper runs, how a dial clicks or how a liquid pours out of your bottle, so the model guesses, and the guess often looks plausible and wrong. For clips that demonstrate function, film the real product.
Fine printed detail is the other limit. Labels, ingredient panels, small logos and model numbers tend to blur or morph during motion. If the legible text matters to the buyer, keep it in the still images and let the clip show shape and finish.
Clips are also short. Models like wan-2.6-i2v go up to 15 seconds and kling-v2.6 to 10, which is a product loop, not a full walkthrough. There is a further boundary that no model can cross for you: the marketplace decides what is allowed in the video and how it is reviewed. People appearing in the clip should be your own or fictional, never a recognizable stranger. Treat the result as a draft of your product shown moving, then check it against your own item and the platform rules.
Can I make an Amazon product video from just my listing photos?
Yes, as a short clip. You upload a listing photo as the first frame of an image-to-video model, describe a slow camera move, and generate five to ten seconds of motion. It works best with sharp photos on clean backgrounds. It will not replace footage that shows the product being used, but it can give the listing a moving version of the pictures shoppers already see.
Which AI models turn a product photo into a video?
On Picasso IA, models that take a photo as the starting frame include kling-v2.6 and wan-2.6-i2v, both in the text-to-video category of the catalog. Kling offers 5 or 10 second clips and wan-2.6-i2v offers 5, 10 or 15 seconds with 720p or 1080p output. Run the same photo through two of them and keep the take that holds the product most accurately.
How do I keep the product looking exactly like the real one?
Start from a sharp photo, describe only camera and light movement, keep the motion slow, and compare every take with your listing images before using it. Reject any take where a shape, color, button or label changes. The model starts from your photo, but it still redraws each frame, so the review step is not optional. If detail keeps drifting, shorten the move or try another engine.
What are the current Amazon video requirements?
They are set by the marketplace and they change, so this page does not quote them. Length, file size, format, aspect ratio and content rules vary by category and by account type. Open the seller help pages for your listing category and read the current video requirements before you generate, so you can pick the right duration and framing from the start.
How long can the clip be?
It depends on the model. kling-v2.6 offers 5 or 10 seconds, and wan-2.6-i2v offers 5, 10 or 15 seconds. That is enough for a loop of the main product shot or a detail close-up. If you need a longer video, plan it as several short clips that you edit together, each starting from a different listing photo.
Can the clip include sound?
Some models can. kling-v2.6 generates audio by default and has a switch to turn it off, and wan-2.6-i2v accepts an audio file to sync with the clip. For product clips, many sellers prefer to switch generated audio off and add a plain music track afterward, so check what the marketplace allows and what suits your listing.
Can I put a person in the video?
Only a person who is yours to show: yourself, someone who agreed, or a clearly fictional character. Do not generate a clip that looks like a real identifiable person without permission. For hands in scale shots, check that fingers keep their shape through the whole clip, because hands are where models slip most often.
Will this make my listing rank higher?
Nobody can promise that, and this page does not. A video can help a shopper understand the product, and that may help the listing, but ranking depends on many things the marketplace does not publish. Use the clip to show the product clearly and accurately, then judge it by how shoppers respond rather than by a promise from a tool.
What if the generated clip changes the product?
Throw that take away and generate again. Try a slower move, a shorter prompt, a stronger negative prompt such as extra objects or changing colors, or a different model. A clip that misrepresents the item can lead to returns and complaints, so a take that is almost right is still the wrong take.
How much does it cost to make a product clip?
Generations are paid in credits, and the cost depends on the model, the duration and the settings you pick, so a number written here would go stale. New accounts get free credits to test a first clip. Current plans are on the pricing page, and video generally costs more than a still image.
Ready to try it on your own listing? Open the Picasso IA toolkit, attach your best product photo and generate the first take.