Google is revolutionizing the shopping experience with the integration of generative AI into Google Shopping. As part of a broader update, Google has unveiled a virtual fitting tool for apparel, utilizing generative AI to capture clothing images on real models in various poses. This innovative technology enables customers to virtually try on clothes and even predicts how shadows will form, enhancing the online shopping experience.
Google’s virtual try-on feature is powered by a new diffusion-based model developed internally. Following the footsteps of text-to-art transformations like Stable Diffusion and DALL-E 2, the Diffusion model gradually subtracts noise from an initial image composed entirely of noise, step by step, to move closer to the desired outcome. This approach, applied to clothing images, allows Google to simulate how clothing drapes, creases, fits, stretches, wrinkles, and how shadows interact with the fabric.
To enhance the model’s robustness and address visual imperfections such as distorted folds, Google utilized pairs of images depicting individuals wearing clothes in different poses. By repeatedly processing random pairs of clothing and people images, Google refined the model, ensuring a more realistic representation of apparel.
Starting today, US Google Shopping users can virtually try on women’s tops from popular brands such as Anthropologie, Everlane, H&M, and LOFT. Users can identify available items for virtual try-on through the new “try on” badge in Google search results. Men’s tops are scheduled to be included later this year, broadening the offering to a wider audience.
Virtual try-on technology is not entirely new, as companies like Amazon, Adobe, and Walmart have also explored generative modelling and real-time clothing rendering. However, Google’s implementation stands out with its emphasis on using real models and catering to diverse body types, races, skin tones, and hair types.
Simultaneously with the virtual try-on feature, Google will implement AI and visual matching algorithms to filter clothing searches within Google Shopping. Users can refine their search results across various stores based on specific criteria such as colours, styles, patterns, and more. This feature aims to replicate the assistance one would receive in physical stores, where associates suggest alternative options based on what customers have already tried on
Google’s integration of generative AI and virtual try-on technology marks a significant advancement in the online shopping landscape. By enabling customers to virtually try on clothes and predicting how the clothing will drape and interact with shadows, Google is enhancing the accuracy and convenience of online apparel shopping.