What Google’s Virtual Try-On Means For The Future Of Fashion

Google unveiled a new virtual try-on feature that enables users to visualize how clothes look with varying body shapes and sizes.

Customers can now select a model to virtually try on clothes from popular brands such as H&M, Loft, Everlane, and Anthropologie. More brands to follow.

A “Try On” badge is added to Google’s shopping search which drives the experience.

“Our new generative AI model can take just one clothing image and accurately reflect how it would drape, fold, cling, stretch and form wrinkles and shadows on a diverse set of real models in various poses,”— Lilian Rincon, Google senior director of product and shopping

Here’s what it looks like on an iPhone:

Google’s Groundbreaking Model For Fashion

As amazing as this is for customer shopping experiences, it’s important to note what Google has accomplished here.

The key accomplishment was taking a photorealistic and detail-oriented visualization of a garment that can adjust to significant body pose and shape changes for different body types.

Google did this via a new diffusion-based AI model referred to as TryOnDiffusion a diffusion-based architecture called Parallel-UNet

credit: https://tryondiffusion.github.io/

Unlike traditional diffusion methods that rely on text input, this innovative approach utilizes a pair of images — a garment image and a person image.

Each image is processed separately by its own neural network, specifically a U-net, and then the networks exchange information through cross-attention.

The result is an output image that accurately depicts the person wearing the garment. This combination of image-based diffusion and cross-attention constitutes the core of this groundbreaking AI model.

The Future Of Fashion Models

In all of the virtual try-on announcements, Google emphasizes that “real” fashion models are utilized.

And that’s fine, it’s a way to keep the focus on the feature.

But what I see is the beginning of a huge disruption in the fashion industry.

If you take Google’s TryOn technology and couple it with highly realistic image diffusion models on the scale of Midjourney, you can easily create AI human fashion models wearing retail clothes in just about everywhere… advertisements, marketing, catalogs, posters, etc…

Levi’s is already planning to use AI clothing models under the guise of diversity and inclusion — referred to as ‘artificial diversity’

credit: Levi’s AI clothing model

The irony of course is instead of using diverse people as the fashion models, they would use AI fashion models to portray the diversity.

Strange move.

It should be noted that Levi’s is not using Google’s tech, but Lalaland.ai, a company that uses AI to create hyper-realistic ‘body-inclusive avatars’.

Considering the sensitivity of AI job replacement, most retailers are a bit cautious at the moment about using AI fashion models. They don’t want any brand customer backlash.

But the cost savings and range of possibilities in using AI fashion models has to be on their mind, and it’s only a matter of time before smaller retailers will start to use AI fashion models forcing larger retailers to compete.

Google has cracked the technology. Now it’s about cultural acceptance.

The Future Of Fashion Shopping

But it’s not just AI fashion models that will impact the fashion industry, what shoppers really want to see is virtual try-on via their own bodies.

Walmart recently released a Be Your Own Model providing online shoppers the ability to upload their own photos and visualize how various clothing items will look on them.

Walmart hasn’t released the technical details of how the technology works and I suppose it is only for brands sold in Walmart.

As Google is integrating the try-on feature into its search capability, it is expected to have a more significant impact on a broader audience.

And once Google allows for customers to use their own body photos, which is likely the next move, this will take Fashion shopping to the next level.

Final Thoughts

Google’s AI for shopping will break significant ground in the fashion industry.

These advancements mark significant steps towards a future where AI fashion models and personalized virtual try-on experiences become commonplace in the world of fashion shopping.

My humble prediction is that in 3 years, the retailer catalogs you get in the mail (if they still exist) will likely be all AI fashion models.

And it won’t matter because no one will be able to tell the difference.