How to Create AI Product Videos with Seedance 2.0 Mini

Learn how to use Seedance 2.0 Mini to prototype short AI product videos, control shots, and review clips before polishing them in CapCut.

*No credit card required
Amber glass bottle on a wooden block, with a tablet camera preview beside it
CapCut
CapCut
Aug 12, 2026

Seedance 2.0 Mini is best used to test short product-video concepts, particularly when you need several simple motion treatments from an approved product image or campaign idea. Its output is limited to 480p or 720p and clips are designed to run up to 15 seconds, so treat it as a route for short-form prototyping rather than an automatic choice for high-resolution hero assets.

The reliable workflow is simple: start with an approved product reference, generate one controlled shot at a time, review the output against the real SKU, then edit only approved clips into a finished channel asset.

Hand placing a small product photo on the first of three blank taped cards

Decide where the video will appear before you generate it. That choice determines framing, the product detail that needs to be visible, and whether a 480p or 720p source is appropriate for the final delivery requirement.

Seedance 2.0 Mini can support text-to-video, image-to-video, and reference-guided workflows. Available controls differ by access route, however, so confirm the current official documentation for the interface you are using before planning around a particular duration, aspect ratio, upload rule, or generation mode.

Use the mode that matches the level of product control you need:

Table comparing text-to-video, image-to-video, and reference-guided generation use cases

For product marketing, image-led generation is usually the safer starting point when packaging, finish, or color must resemble the approved product. It does not guarantee an accurate result; it simply gives the generation a clearer visual source.

Choose the destination format at the beginning. Mini aspect-ratio options can include 16:9, 9:16, 1:1, 4:3, 3:4, and 21:9, but not every route exposes every option.

    1
  1. 9:16: vertical social placements.
  2. 2
  3. 16:9: horizontal video placements.
  4. 3
  5. 1:1 or 3:4: compact product-focused social or catalog-style treatments, if supported by your route.

Avoid generating a wide scene and hoping a later crop will preserve the product, hand interaction, and text-safe space. Create the right aspect-ratio version from the outset.

A strong reference package prevents avoidable drift before you write a single prompt. For image-to-video, the upload is part of the input alongside the prompt and available output settings.

Prepare:

    1
  1. One clean, well-lit, uncropped, uncluttered product image.
  2. 2
  3. An approved alternate angle, if it is needed for the shot.
  4. 3
  5. Brand-approved logo files and color references for later editing.
  6. 4
  7. A short list of non-negotiable attributes: SKU, pack shape, cap or closure, label placement, colorway, material, and visible product claims.
  8. 5
  9. A separate list of what may be invented creatively, such as the background, lighting mood, surface, or non-product props.
  10. 6
  11. Required legal copy, claim substantiation, and product or brand approvers.

If you use multiple references, give each one a job. For example: "Image 1 is the product's approved front label. Image 2 is the reference for the matte bottle finish. Do not use Image 2 as label artwork." Clear roles reduce ambiguity, but they do not eliminate altered logo treatments, unreadable packaging text, incorrect reflections, or fine-detail errors.

Treat the reference image as the source of truth during review-not as proof that the generated video is accurate.

A short product video is easier to control when it is made from discrete clips rather than one overloaded sequence. Plan a small library of shots that can be assembled later:

    1
  1. Hero reveal: establish the product and its silhouette.
  2. 2
  3. Detail shot: show one material, texture, or feature.
  4. 3
  5. Use-context shot: place the product in a simple, relevant environment.
  6. 4
  7. Benefit moment: show a visual demonstration only if it can be reviewed accurately.
  8. 5
  9. End-frame space: leave room for approved copy, logo, and CTA added in editing.

Each generation should have one product, one message, one primary action, one environment, and one camera behavior. This reduces the risk of the model changing the SKU while trying to satisfy competing instructions.

A useful prompt structure is:

Approved product/reference + product action + visible detail + camera movement + setting + lighting + mood + ending frame

For prompt-writing guidance on structuring the subject, action, camera, environment, and mood, see Seedance 2.0 Mini prompt guidance.

Use the uploaded approved bottle as the product reference. Preserve its cylindrical shape, matte black finish, white front label placement, and gold cap. The bottle rotates slowly on a dark stone surface. The camera makes a slow left-to-right orbit at close product distance. Soft side lighting creates a restrained reflection and a natural shadow. Minimal premium studio mood. End with the bottle centered and empty space above it for approved copy.

The important distinction is between product movement and camera movement:

    1
  1. Product movement: "the bottle rotates slowly."
  2. 2
  3. Camera movement: "the camera makes a slow left-to-right orbit."

Keep them separate. A vague instruction such as "cinematic dynamic motion" leaves too many choices open.

You can also specify material, reflection behavior, brand colors, motion speed, and camera direction, but these instructions are not guarantees of accurate rendering. CapCut's product-video guidance recommends describing those visual factors directly rather than relying on broad style labels.

Do not make generated packaging copy, a precisely redrawn logo, regulated claims, price, or legal language central to the shot. Add verified text later in the edit.

Man comparing printed product shots on a wall and pointing with a red pencil

Generate a small set of variations, then review each clip in the same order. Do not start by judging whether it looks impressive. First determine whether it is safe and accurate to use.

    1
  1. Does the visible product match the approved SKU?
  2. 2
  3. Are the shape, color, packaging, closure, and key material cues correct?
  4. 3
  5. Is the logo treatment acceptable, and is any visible text accurate enough to retain?
  6. 4
  7. Has the clip introduced an unsupported claim or misleading product behavior?
  8. 5
  9. Are hands, object contact, shadows, and reflections plausible?
  10. 6
  11. Is the framing usable for the intended crop and overlay?
  12. 7
  13. Does the motion support the intended message?

Reject a clip if it changes the product, presents a misleading package, or creates an unapproved claim. A visually attractive result is not a usable marketing asset if the SKU is wrong.

Hands, object contact, crowded scenes, fast movement, and complex interactions can require retries. Reflective materials and tightly detailed labels also deserve close review. When a result is close, revise methodically: change one variable at a time.

For example:

    1
  1. If the bottle is accurate but the shot feels flat, retain the product reference and action; revise only the lighting instruction.
  2. 2
  3. If the camera is too active, retain the setting and product description; simplify the camera move.
  4. 3
  5. If a hand changes shape or fails to contact the object convincingly, remove the interaction and generate a product-only shot instead.
  6. 4
  7. If the label drifts, return to the approved reference and reduce scene complexity rather than adding more competing instructions.

Where your interface offers a seed control, retaining the same seed, shot size, and motion path can make prompt comparisons cleaner. Do not assume every Seedance route exposes that control or produces deterministic repeats.

Once a clip passes product and brand review, bring it into CapCut for the production work that should not depend on generated text or perfect in-model composition.

Use the edit to:

    1
  1. Trim each clip to its strongest moment.
  2. 2
  3. Arrange a clear reveal-detail-benefit sequence.
  4. 3
  5. Add approved product copy, captions, pricing, legal language, and CTA.
  6. 4
  7. Apply approved logo and brand elements.
  8. 5
  9. Add audio you own or are licensed to use.
  10. 6
  11. Create vertical and horizontal versions from the approved source material.
  12. 7
  13. Check readability, safe framing, and product visibility in the final destination format.

Do not treat generation as a blanket commercial-use license. Before publishing, confirm that your team has rights to every input asset and that the exact Seedance access route, account, region, and platform terms permit the intended use. Also clear likeness permissions, trademarks, applicable advertising rules, and any required disclosure of synthetic media under your brand policy, platform rules, or law.

After the selected Seedance clips have passed product, legal, and brand approval, open CapCut to build the final 6-15 second sequence. Add approved copy and branding, make the vertical and horizontal versions your campaign needs, and export the version that matches its publishing destination.

Hot and trending