by 
One of the ongoing questions and hot topics in AR is how it’s converging with AI. We wrote an entire white paper on it, and were on stage at AWE USA 2024 when Snap announced GenAI suite: a new set of generative AI tools meant to automate and assist AR lens creation.
AR creation workflows are just one of the collision points between AI and AR. Another one will be what we call “generative AR,” which is serendipitous AR experiences that are prompted by users on the fly. This deviates from the highly pre-ordained and produced flavors of AR to date.
Another convergence of these technologies recently crossed our desks: eCommerce. Generative AI is increasingly integrated into eCommerce engines to let retailers and merchants generate their product listings through text prompts, thus saving time to produce product images.
This falls into the broader AR camp in that media is being manipulated in value-added ways. This doesn’t necessarily involve 3D images or AR product try-ons (though it could) but rather 2D eCommerce. AR is present in the broader sense that media is being augmented.
With that backdrop – and in honor of Black Friday – here are the latest examples of generative AI in eCommerce…
Amazon
Taking those one at a time, Amazon was first out of the gate with its AI image generation tool. This helps non-media savvy sellers generate professional-looking images for their product pages or ads. Specifically, the primary function is generating contextually relevant backgrounds.
This works similarly to popular flavors of generative AI in that merchants can enter keyword prompts to produce images. They first upload product photos, then type image descriptions or themes (think: beach, kitchen, autumn, sunset, etc.). It then spits out snazzy backgrounds.
Merchants can play with this functionality until they get the image they want – a process that will naturally condition their skills as prompt engineers. But it’s meant for any skill level, and will apply to small and large eCommerce sellers that can benefit from time and cost savings.
Moreover, this addresses a broader area that’s primed for AI: brand marketing. Early in the generative AI cycle, this was the area we pointed to as the most prone to AI disruption, along with stock photography. Generating images on the fly scratches an itch for brand marketers.
The 3D Dilemma: Understanding Leadership Hesitation in Fashion
Moving on to Google’s latest integration, it similarly generates custom images for product pages, such as Google Shopping. Known as AI-Powered Product Studio, it features text-to-image AI that’s driven by prompts. And like Amazon’s play, its primary purpose is image backgrounds.
The use-case example that Google spotlights is a beauty brand that can position skincare products surrounded by seasonal imagery or ingredient-evoking fare like peaches and tropical plants. It’s all about boosting the “craveablity” of a given product and, ultimately, conversions.
In addition to adding backgrounds, Product Studio can do things like remove distracting backgrounds and replace them with simple color gradients or other modern looks. It can also improve and upscale low-quality images, which could save brands money by avoiding reshoots.
Speaking of avoiding reshoots, one value proposition behind Product Studio (and Amazon’s corresponding play) is to engender more variety from a single product photo. That includes seasonal variations or more specific thematic elements like holidays or moods.
Meta
Meta continues its land grab for AI. Joining the Meta AI suite of generative AI products (images, text, etc.), the company is now expanding its image generation capabilities for advertisers. This is meant to reduce friction in formulating ads on Facebook with the help of genAI for product visuals.
For example, a coffee shop can generate an image of a logo-emblazoned coffee cup from several angles. It can also create variations and thematic touches like the cup sitting on top of a pile of coffee beans, or other branded fare in the background, such as the shop’s location.
Though this latest update is all about images, Meta reminds us that it can also generate other campaign components such as ad copy and headlines. Next up on Meta’s road map is to emulate a given SMB’s personality and voice, using previous campaigns as a training set.
Meta says that all the above will be available to all advertisers by the end of the year. Meanwhile, it’s already working. In early trials, smartphone case maker Casetify boosted its return on ad spend (ROAS) by 13 percent by streamlining the creative process in these ways.
eBay
The list of eCommerce players integrating generative AI continues to grow. Joining the above players is one of the original eCommerce engines: eBay. And like the above examples, its AI integrations so far are all about generating product images and streamlining product listings.
Specifically, eBay offers a drag-and-drop tool to replace image backgrounds. Powered by Stable Diffusion models, this involves removing and replacing backgrounds to be more colorful, seasonally relevant, on-brand, or thematically aligned with the product itself.
As you can imagine, this democratizes advanced product imagery, as most sellers are only equipped with their smartphone camera. In the end, eBay’s genAI tools can help small sellers gain production quality and customization previously reserved for pros and large brands.
This notably follows a progression of eBay moves with AI, such as generative AI product titles and descriptions. That elevated SMB sellers’ abilities to differentiate their listings through text, while the latest move brings generative images into the mix – a nice one-two punch.
Point of Integration
So there you have it. Expect to see more development on this front as eCommerce is a natural point of integration for generative AI. This will materialize on the merchant end and the consumer end. For the latter, conversational AI interfaces will help users find the right products.
More from AR Insider…
