How-to guide

AI Video: Start a Product Clip from Reference Photos

Add product reference photos, describe the scene, and review the finished clip against the details you supplied.

9 min read · AIVIDEOSUGC

Product reference photosScene descriptionReview the finished clip

AI Video workspace showing reference upload, prompt, and settings controls.

AI Video is for a clip built from product references and a written scene direction. It starts from visual material you provide; it is different from Text to Video, which starts from a brief, and Image to Video, which starts from one photo and one motion prompt.

The yoga-mat imagery below is a homepage example. The short clip is a separate silent sample with captions already in the footage; it is not represented as the output of this tutorial run.

Product reference → scene direction → visual sample

The teal mat photo is the product reference. The gym presenter clip is a separate homepage sample for framing; it was not generated from this photo.

Product reference · exampleTeal yoga mat product photo on a plain background.Homepage fitness sample · separate from the mat example
Example direction to try: show the mat unrolling on a clear studio floor; keep its teal color and edges recognizable.

Build a clip from product references

The exact controls depend on the available model. Follow the options shown in your workspace.

  1. Open AI Video

    Go to [AI Video](/app/create/ai-video-generation). If the workspace asks you to sign in, return to AI Video after signing in.

  2. Add clear product references

    Upload the photos that show the product clearly from the angles you want the model to use. Follow the current tool’s file-count and format limits; remove unrelated images.

  3. Describe one scene direction

    Say what should happen, where it happens, and which product details must stay visible. Keep the example focused on the mat unrolling and retaining its teal color.

  4. Review the settings and estimate

    Check the selected model, ratio, duration, and other controls on screen. Read the estimate before you start. If a control or model is unavailable, use one that the current workspace offers and update the example settings accordingly.

  5. Generate and check the clip

    Start the job and wait for completion. Compare the product silhouette, color, label, and action with the reference photos. Look for changes between shots before keeping the result.

  6. Handle a failed or inconsistent result

    For an upload error, follow the message shown for that file or model. If the product changes shape or the action is unclear, simplify the scene description and change one detail before retrying.

Checklist

  • Every reference photo shows the intended product.
  • The scene description names a visible action and key product detail.
  • The chosen settings match what the workspace actually offers.
  • The finished clip remains recognizable as the referenced product.

Related

FAQ

Questions

How is AI Video different from Image to Video?

AI Video uses product references with a broader scene description. Image to Video starts with one photo and a specific motion prompt.

Can every model use the same references and settings?

No. Check the file limits and settings shown for the model selected in your workspace.

Is the yoga-mat clip generated by this walkthrough?

No. It is a separate homepage sample, and its captions are part of the footage.

Open AI Video

Add product references, describe the scene, and review the estimate before generation.