Publisher disclosure: This guide is published on behalf of the operator of seedanceai.im. SeedanceAI is an independent third-party tool, not affiliated with ByteDance or Seedance. Product details were checked on November 12, 2026. The example below is a proposed production brief, not a claim that we generated or tested a finished video.

A product photograph can be an excellent starting point for a short video. It already settles several decisions: the object, its material, its proportions, and the lighting. The challenge is deciding what should change without losing those details.

For a first attempt, make the job smaller. Choose one product, one action, and one camera move. A five-second clip that communicates a single idea is easier to inspect than a miniature commercial with multiple locations, transitions, and competing instructions.

1. Write the acceptance criteria before the prompt

Consider an unbranded ceramic mug on a plain tabletop. The objective is a short product reveal, not a transformation. Before uploading anything, write down what would make the shot usable:

  • The mug keeps the same shape and handle throughout the shot.

  • The camera movement makes the product easier to see.

  • The background stays unobtrusive, with room for a caption added later.

  • The last frame can hold briefly without an awkward cut.

These criteria give you a practical rejection list. Attractive lighting does not compensate for a handle that changes shape. A dramatic camera move is not useful if it hides the product.

2. Choose an input mode for the job

The SeedanceAI workspace presents Text to Video, Image to Video, and Multi Reference modes. For a product shot, use Image to Video when you already have the composition you want. The site describes a first-frame workflow with an optional last frame. Use Multi Reference when separate assets need to guide different aspects of the shot.

Do not add references simply because the upload limit permits them. Give each file a job. One photograph can establish the product; a short clip can demonstrate a camera path. If two images show incompatible lighting or different object proportions, decide which should govern before asking the model to reconcile them.

Use photographs, footage, and audio you own or have permission to use. A publicly visible clip is not automatically licensed for reuse. Remove private information from backgrounds and obtain appropriate permission before using identifiable people.

3. Describe the change, not every visible detail

Here is an illustrative brief for a single product photograph:

Keep the ceramic mug and tabletop consistent with the starting image. Over five seconds, make a slow, small push toward the mug. The mug remains stationary and fully visible. Keep the lighting soft and steady, and finish with a clear view of the handle. No scene change or new objects.

This is a starting hypothesis, not a guaranteed recipe. The important choices are explicit: the subject stays still, the camera moves, and the shot does not cut. If the result misses the brief, change one instruction at a time. For example, reduce the camera movement before adding more adjectives about realism.

In Multi Reference mode, the site provides an @ reference mechanism. Use it to specify the role of an uploaded file instead of assuming the model will infer whether you intended its subject, movement, or sound.

4. Budget for decisions, not just duration

The published credit table lists a five-second, 720p text-to-video clip without video or audio references at 15 credits for Seedance 2.0 and 20 for Seedance 2.5. Three attempts at those exact settings therefore total 45 or 60 credits. This arithmetic is a planning example, not a price quote for the image-to-video brief above.

Check the actual quote in the generator. Reference media, duration, and resolution can change the total. The pricing page offers subscriptions and one-time packs; the smallest listed one-time pack is $9.90 for 60 credits. Generating requires credits, and registration does not guarantee a free grant.

Set a retry limit before the first run. A useful small experiment is a baseline, one targeted revision, and a final comparison. Do not simultaneously change the model, prompt, reference image, and resolution: you will lose the ability to identify which change helped.

5. Review at normal speed and frame by frame

Watch once without pausing. Does the viewer understand the product immediately? Then inspect the opening, middle, and closing frames. Check silhouette, handle geometry, reflections, contact with the tabletop, and background stability. If sound is present, review it separately rather than letting a convincing soundtrack distract from visual defects.

Keep a short log: input files, prompt, model, settings, quoted credits, observed defect, and next change. Label any successful clip with the exact version of the brief. This turns a sequence of retries into a repeatable production process.

Finally, add essential logos, prices, and captions in an editor where you can control the typography. Review the exported file in the intended placement, including a phone-sized preview. Publish only a result that represents the product honestly. The goal is not the most elaborate prompt; it is a clear shot with a defensible acceptance test.