How to turn a product photo into an AI video ad
You have product photos and you want a video ad. Most of the gap between the two is about giving a video model enough to work with, and then checking the result the way a brand manager would, not a fan.
What you need before you start
- A clean hero photo. The product on its own, evenly lit, whole in the frame, label facing the camera. A phone photo next to a window is enough.
- A label close-up. The logo and wording that must stay readable in the ad.
- Other angles, only if you have them. A back or side view helps, but don't ask a model to imagine the back of your packaging and then trust what it draws.
- The facts. Product name, size, what it's for and the exact label text. Mark anything you don't know as unknown.
Why one photo isn't enough
Video models fill gaps. Given one front-on photo, the model has to guess how big the product is in a hand, what the back looks like and how the lid opens. The usual failures follow from that: label text turns into look-alike gibberish, the product grows or shrinks between moments, colours shift under new lighting, or a part appears that your product doesn't have.
A product sheet helps. It lays out the hero view, the props that belong with the product and a close-up of the label on one reference image, so the model has the supplied views and details together. A generated sheet does not establish unseen angles or unknown dimensions; mark those as not supplied and check the sheet against your originals. Keep the original label close-up as its own reference too, because that's the detail most likely to drift.
Add a person and a place
Choose whether a person helps show your product's use. A product-only demonstration can work too; test both if your budget allows. Give the model a character reference (a front portrait and the same person from several angles, in one outfit) and a scene reference that matches where your customer would really use the product: a bathroom shelf for skincare, a desk for a gadget, a car cup holder for a drink.
Direct one clear action
Pick a single thing the person does with the product: apply it, open it, take the first bite, pour it, put it on. Then describe the contact between hand and product precisely. "She holds the jar in her left hand, label to camera, and dips two fingers in" gives the model far less to invent than "she uses the cream".
In this five-second example the action is one sweep of blush, and the open compact stays in her other hand, turned toward the lens, the whole time:
State the scale when you know it; never invent a measurement. "A 50 ml jar that fits in her palm" stops the product from turning into a tub. If you want the creator to speak, write the exact line and keep it short; there's room for roughly one sentence in ten seconds. The AI video ad prompt template shows how to lay all of this out.
Check before you post
Watch the generated video with the sound on, then again frame by frame around the moments the product is closest to the camera.
- Label: the text and logo match your product exactly, with no extra words.
- Shape and scale: the product is the same size in every moment, and the right size next to the hand.
- Colour: it matches your photo, allowing for the scene's light.
- Hands: the right number of fingers, and a grip that makes sense.
- Speech: exactly the approved line, in sync, with no added claims.
- Claims: nothing on screen or in the dialogue that the product can't back up.
- Disclosure: the ad is labelled as AI-generated.
If something fails, fix the input that caused it. A wrong label usually means the label close-up is missing or blurry; a product that changes size may need a confirmed measurement or a real scale reference.
From product photo to video ad in fairytale video
In fairytale video you upload clear product photos and a label or detail close-up. Review the originals before generating. The studio prepares a product sheet when the video needs one, with its cost included in the Generate quote. Choose a creator from the library or create your own, pick a scene, and describe the ad and exact spoken lines. You can also leave out the creator for a product-only ad. Choose 9:16 or 16:9, 4 to 30 seconds, and 720p or 1080p. The Generate button shows the price of the Seedance 2.5 render before you start. Each video writes a concept and timed plan and stores the exact prompt and references it used, preserving that history after later edits. Check it, add captions at no extra credit cost, and download it. More on the format in AI UGC ads: creator-style video ads without a shoot.