Product Video Prompts That Hold Product Identity

Most bad AI product footage fails for one reason: the product changes. The label shifts, the lid swaps sides, the colour drifts two shades between frames. A product video prompt is the tool for preventing that — not by describing the product more beautifully, but by describing the shot so precisely that the model has no room to reinterpret what it is looking at.

Why products are unforgiving subjects

• Identity is the whole point. A viewer comparing your video to the product on a shelf is checking shape, colour, and label position. Drift is a defect, not a style.
• Reflective surfaces. Bottles, cans, screens, and packaging all mirror their environment. Unspecified reflections read as smudges.
• Fine print does not generate well. Ingredients lists, model numbers, and legal text come out as plausible-looking noise.
• Hands enter the frame. As soon as someone holds the product, hand anatomy becomes part of the shot.
• Studio light is a convention. Product footage has recognisable lighting grammar; breaking it makes the clip feel like stock rather than brand.

The prompt slots for product shots

1. Staging — where the product sits: seamless backdrop, textured stone, wooden desk, held in frame by a hand, shelf with competing objects.
2. Product and finish — shape, material, colour, matte or gloss, condition. Describe it; do not name a real brand you do not own.
3. Hero action — the one thing that happens: a slow rotation, a pour, a cap being lifted, a cloth wiping the surface, a lid opening.
4. Lighting — key light direction, softbox softness, rim light to separate the edge, gradient background.
5. Camera — slow orbit, locked-off macro, gentle push-in, overhead flat-lay with movement. One move.
6. Environment cues — water droplets, steam, dust, fabric texture: the small physical evidence that sells realism.
7. Constraints — keep label position, colour, and shape stable; no on-screen text; no extra products entering frame.

Copy-ready examples

Slow rotation on a seamless backdrop

A matte ceramic skincare bottle rotates slowly on a seamless pale backdrop, soft top-left key light with a warm rim light, label area kept free of text, camera locked off at macro distance, single 6-second rotation, product colour and shape stable throughout.

Pour shot

A dark glass bottle pours a translucent amber liquid into a clear tumbler on a wooden counter, low side light, slow motion feel, camera fixed at table height, steam-free, no hands in frame, liquid flow smooth and continuous.

Held in hand

A compact grey device is held steady in one hand against a blurred workshop background, product facing the camera, gentle handheld sway, cool daylight from a window, thumb resting naturally on the lower edge, no text or logos visible.

Flat-lay with motion

An overhead flat-lay of three unbranded skincare containers on pale linen, the middle container slides slightly toward the camera, soft even light, static overhead camera, fabric texture visible, arrangement stays orderly and the containers do not change shape.

Generate a product shot in three steps

1. Anchor the product before anything else. Open the LongCat Video generator in image-to-video mode and upload your product photo. This is the single highest-leverage step for identity: the model starts from your actual packaging instead of re-imagining it.
2. Prompt only the motion and light. Since the model can already see the product, spend the prompt on the hero action, the light direction, and the camera move. Re-describing the image wastes the prompt on information the model already has.
3. Extend for longer demos. Use continuation to grow a shot when a demo needs more screen time than one clip, then cut in your editor. The image to long video workflow covers holding one subject across a longer sequence.

When this is the wrong approach

• Legible on-screen text. Prices, taglines, and legal disclaimers have to be added in an editor. Generated lettering is not usable, and the prompt cannot fix it.
• Accurate label reproduction. If the packaging carries fine detail that must match print exactly, shoot it or composite a real product plate over generated footage.
• Multi-product comparisons. Three products rotating independently in one frame is where coherence goes first. Generate one product per shot and edit them together.
• Trademarked brands you do not own. Do not prompt another company's packaging or logo as your subject — see the acceptable use policy.

Common mistakes

• Writing a product description instead of a shot description. "Our serum is luxurious and lightweight" is copywriting, not prompting. Say what moves and how the light falls.
• Generating text in the frame. If the label must read something, leave the label visually empty in the prompt and add the copy in post.
• Changing two things at once. Rotation plus a push-in plus a light change guarantees drift. Change one variable per take.
• Forgetting the surface. Products float without a shadow contact point; name the surface and the contact shadow.
• Ignoring reflections on glossy packaging. State what the environment reflects, or the model will invent a distracting room.

Related pages

• Consistent character AI video — the same identity-hold method for faces and products over time
• Car video prompts — the vehicle version of these seven slots
• Dance video prompts — when people, not products, are the subject
• AI video motion control — controlling rotation speed and camera arcs precisely
• LongCat Video pricing — estimate credits before a batch of product takes