2026-09-28

Product video prompt examples: a template and 6 shots

A product video prompt describes one shot: the action, one camera move, the light, the setting. A template, six examples, and what goes in fields.

5 min readSume

All posts

A good product video prompt describes one shot in plain sentences: what happens to the product, one camera move, the light, and the setting. Send the packshot as the clip's first frame so the image, not the words, sets how the product looks, and put the length, shape, and sound in the request's fields.

Sume's Video generation docs, read on 2026-09-28, ask for the same things: “details about motion, camera angles, lighting, and scene composition”. Their own product example is one line, “A vertical UGC-style product clip on a desk, natural light”, with the shape and length set in fields. The six prompts below are our own examples, written as starting points rather than tested recipes.

What should a product video prompt include?

Five parts, one short sentence each. Name the action (a pour, a lid lifting, a hand picking the product up), then one camera move or a still camera, then the light, then the surface and background, and last what must stay put, such as “the label faces the camera”. One move per clip keeps the shot readable; camera movement prompts explains the terms. As a template:

The {product} {action}. {One camera move, or "the camera stays still"}.
{Light}. {Surface and background}. {What must not change}.

Should I describe the product or use its photo?

Use the photo. Send the packshot as the clip's first_frame, which sets how the clip starts, rather than describing the product or attaching it as an input_references guide; Product logo warping in image-to-video explains the difference and how to check the label afterwards. Then spend the prompt on what changes rather than on the product's look, and don't ask for views the photo can't show, such as a full turn from one front-facing packshot.

What are some product video prompt examples?

Each prompt covers one shot. Swap in your product and keep the structure:

Prompts are examples written for this post. Frame rules from Video generation, read 2026-09-28.
Shot	Prompt	Frames
Push-in on a packshot	Slow push-in toward the serum bottle on a marble counter. Soft morning window light. The bottle stays centered, label facing the camera.	Packshot as `first_frame`
Pour	Golden honey pours from a wooden dipper into the open jar. Slow motion, warm backlight, dark wood table. The camera stays still.	Photo of the jar as `first_frame`
Unboxing	Two hands lift the earbuds case out of the open box and set it on the desk. Top-down view, bright even light, white desk.	Photo of the open box, case inside, as `first_frame`
In use	Steam rises from the coffee as a hand reaches in and lifts the mug beside the laptop. Gentle handheld sway, late-afternoon light through blinds.	Photo of the mug on the desk as `first_frame`
Texture macro	Extreme close-up as a fingertip swirls the surface of the cream. Slow motion, soft diffused light, pale pink background.	Close-up of the open jar as `first_frame`
Detail-to-product reveal	The camera pulls back from the stitching to show the whole bag on a shelf. Soft studio light.	Detail shot as `first_frame`, packshot as `last_frame`

Which settings go in fields instead of the prompt?

Length, shape, resolution, and sound are the request fields duration, aspect_ratio, resolution, and generate_audio, so leave them out of the prompt; How to write a prompt for an AI video generator lists every field. Two rules matter for product shots: a last_frame, as in the reveal above, works only on models whose supported_frame_images lists it, and there is no negative-prompt field (AI video negative prompt), so describe what must stay put rather than what you don't want.

How do I write a Seedance product video prompt?

The same way. The Seedance models take the same prompt and fields as every model on POST /v1/videos, and the docs publish no Seedance-only syntax. Pin seedance-2.5 for 4–30 seconds, or a Seedance 2.0 id for up to 15:

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: serum-push-in-v1" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "Slow push-in toward the serum bottle on a marble counter. Soft morning window light. The bottle stays centered, label facing the camera.",
    "frame_images": [
      { "type": "image_url", "image_url": { "url": "https://example.com/serum-packshot.png" }, "frame_type": "first_frame" }
    ],
    "aspect_ratio": "9:16",
    "resolution": "720p",
    "duration": 6
  }'

What if I'd rather not write prompts?

Call a catalog Format instead. sume-product-commercial lists “launch films, ecommerce hero videos, product teasers”, and sume-product-usage-demo lists “unbox-and-use clips”. You send the packshot and a brief, and the Format returns a finished video; Ready-made Formats for product video compares them.