All insights

Luma Ray2 and Kling 2: check the product through movement (2026)

Evaluate Luma Ray2 and Kling 2 by checking labels, shape and handling throughout the clip. Record attempts and revisions.

Kevin Dosanjh
Co-founder / CMO, AI Vidia · Updated
Editorial overhead flat lay of two matte product cards on a warm off-white Nordic studio surface representing competing AI video models for product video
On this page7 sections
  1. 01Choose a product that reveals defects
  2. 02Identify what moves
  3. 03A hypothetical perfume-bottle scene
  4. 04Apply the same final requirements
  5. 05Record the path to approval
  6. 06Inspect moments when the product is obscured
  7. 07Keep the conclusion specific

A product video must preserve the item throughout movement. When comparing Luma Ray2 and Kling 2, examine continuity, handling and the revisions needed to produce a usable video.

Choose a product that reveals defects

A neutral object without text may look attractive easily. Your actual product may have a label, a particular finish and recognisable details. Use the item in the brief and define its essential features.

Specify the necessary angles too. An accurate rear view is irrelevant to an ad that only shows the front. If customers need to see the product opening, review the mechanism rather than only the opening still.

Identify what moves

State whether the camera, product or both should move. A simple rotation differs from a hand opening packaging while the camera moves closer. Begin with a limited action so the acceptance criteria are clear.

If that action works, test the more demanding scene and preserve the change in the brief. Otherwise, it becomes difficult to tell whether a defect relates to the product, movement or new camera instruction.

A hypothetical perfume-bottle scene

You want a bottle turned slightly in someone's hand. Approval depends on label text, cap shape, glass colour and liquid level. Check the beginning, middle and end, especially where the hand covers the product.

If the label changes after the hand passes, the clip fails even when the opening looks correct. Simplify the action, try a different reference or retain real product footage. Choosing a different method for a particular scene is a production decision, not a failure of the project.

Apply the same final requirements

CheckQuestion
LabelIs the text accurate and stable?
ShapeDo bottle and cap retain their proportions?
MaterialDo glass and liquid remain consistent?
HandlingDoes the hand make understandable contact with the item?
EditingCan the clip be used in the planned ad?

Assess both samples at the same viewing size and final format. Keep the defects that made other attempts unusable.

Need new ads?
Describe your product and what you want to test.
Send your brief

Record the path to approval

Note version, reference, settings, attempts and editing. A generated clip is not a finished ad. Audio, copy, cutting and export may happen outside the model.

Use Luma's Ray2 guide for the named version and check the actual Kling access being used. Confirm prices and usage terms on the relevant account.

Inspect moments when the product is obscured

Mark the moment fingers pass across the bottle's label. Compare letters, edges and proportions before and after. For recurring defects, test a hand position that leaves the label visible or change the camera angle.

Both methods can be assessed against this limited revision. Keep the original and revised samples so the decision is understandable. If the required action still fails, use real footage for that part. The choice should follow the final product depiction rather than a requirement to generate the entire film with one tool.

For a neutral review, the product owner can examine samples before learning which model made them. Ask for specific differences from the real item, rather than guesses about production technique. Review those defects with the producer and select the method that delivers correct handling and a clear sequence.

Keep the conclusion specific

A useful conclusion might state that one method produced the required rotation within your standards, while opening the product still requires filming. This is more precise than declaring a universal winner.

Repeat a representative scene before basing a larger series on the method. One successful file shows a possibility, not necessarily reliable delivery. Advertising results require a separate actual test. See the guide to combining AI and real product imagery.

Frequently asked questions

01How do we detect product changes?
Compare the reference and video throughout the clip, including before and after a hand or object covers the product.
02Is an attractive opening frame enough?
No. The item must remain accurate in motion and the clip must work in the final edit.
03Can one trial select a model for everything?
A limited product trial supports decisions about the tested assignment. It does not establish a universal winner.

Sources

  1. 01Luma: generating with Ray2

Next step

Get a proposal for your next ads.

Tell us about your product, audience and ad production needs. We review your brief and reply with a proposed next step.

Send your brief

Read next