Amazon’s ad platform now lets vendors upload one product image and automatically generate up to six photorealistic multi-scene video options using the company’s proprietary AI models. The tool targets brands using Amazon’s marketplace and ad network, particularly smaller retailers that previously needed large budgets, production crews and agencies to produce video campaigns. The tool sits within Amazon Ads in Australia and New Zealand alongside existing display and sponsored product formats.
Amazon’s models analyse the original product image and text metadata, then construct short multi-scene clips that place the item in different lifestyle or studio settings. Retailers can cycle through up to six suggested versions, fine-tune scenes and pick the asset that best fits their campaign goals before pushing it live. The platform emphasises photorealism, aiming to make AI-generated clips indistinguishable from footage shot on traditional cameras. Amazon positions this as especially valuable for marketplace sellers that lack in-house creative teams but still need video to compete in crowded search results and social feeds.
That realism creates a new tension. As AI-driven visuals become standard in advertising, shoppers increasingly question whether what they see on screen is a faithful representation of what arrives in the box.
Amazon acknowledges that risk across its $US2.84 trillion ($3.96 trillion) global business and stresses built-in guardrails to prevent “idealised synthetic versions” that mislead customers. The company presents these controls as a deliberate contrast with Chinese-owned marketplace rival AliExpress, where buyers often complain it is hard to tell if the product pictured in a listing actually matches what gets delivered. Accuracy, not just eye-catching content, is emerging as the make-or-break factor for trust in AI-generated retail ads.
