Google expands AI-powered virtual try on feature to include dresses

In its efforts to develop new ways to accurately depict online shopping items, Google has expanded the ‘Virtual Try On feature’, originally launched for men’s and women’s tops. The feature uses AI to display how items look on various body types, considering factors like shadows, wrinkles and draping.

The feature is now being expanded to dresses, allowing users to visualise how dresses appear on various body types. In its official blog post, the company said, “If you’re in the U.S., simply search for dresses on search and click any style that includes a ‘try on’ badge. From there, you can see what that garment looks like on a diverse set of real models, ranging in sizes from XXS-XXXL. Select a model that resonates with you to get a better sense of how the dress looks. Once you find what you’re looking for, click out to the retailer’s site to buy it.”

This feature is powered by a generative AI technology specifically developed for virtual try-on (VTO). The technology utilises a diffusion-based technique that allows generating every pixel from scratch, resulting in high-quality, realistic images of tops and blouses on models. 

However, when testing the diffusion technique for dresses, the company encountered two distinct challenges. First, dresses tend to be more nuanced garments, requiring greater attention to detail. Second, dresses typically cover more of the human body, adding complexity to accurately rendering them in a virtual setting.

This update is expected to enhance the online shopping experience by providing a more realistic view of items before purchase. Although virtual try-on tools improve online shopping, in-store shopping will still be preferred by many for certain purchases. The feature will be available in the United States for now with a possibility of expansion in the future.