A product ad needs accurate representation and a clear message. Compare Sora and Veo on that basis. Cinematic style cannot replace a useful demonstration or supported facts.
Use a brief with explicit exclusions
Choose one variant and message. Protect shape, colour and actual operation. Agree where original imagery or editing is allowed, avoiding an artificial all-AI requirement.
Keep attempts and rejection reasons. Assess first usable output and revision effort separately; a selected success can conceal many failures.
Separate production and advertising tests
An accepted video establishes production under recorded conditions, not superior sales. That needs a relevant advertising test with defined outcomes. Report the two assessments separately.
Verify the chosen service at production time and consider simpler methods where they explain the product better.
Test a specific product problem
Choose a difficult explanation, such as assembly or included items, and define it accurately from original evidence.
Check correct parts, sequence and product/accessory distinction separately. Visual quality must not conceal a failed functional requirement.
Compare a simpler original close-up with text against a generated action where useful. Buyer understanding matters more than technical spectacle.
If launched, record production method with message and offer. Limit conclusions to the actual comparison rather than attributing every purchase difference to the model.
Make protected product details explicit in the brief: the actual variant, package shape, colour and operation. Agree where original photography and manual finishing are allowed so the test reflects a realistic production rather than an artificial requirement to generate everything in one step.
Keep the rejected attempts and their reasons. One selected success can conceal substantial work, while a realistic revision can reveal whether the accepted details survive changes. Separate the time to first usable output from later correction effort. The production conclusion should identify a workable method for that task; an advertising result requires a separate comparison with defined outcomes and conditions.
