Google Doppl App Emerges as a Surprise Breakout Success

I virtually tried on five different outfits in under 10 minutes — or rather, my AI-generated double did. This was all possible thanks to Doppl, a new app currently being tested by Google, which I used to create AI video clips of myself modeling clothing I discovered online. While the app functions reasonably well, it clearly struggles with accurately superimposing pants over photos where I'm wearing shorts. In one particularly amusing case, it even replaced my mismatched socks with artificially generated feet.

Using the app is incredibly straightforward. All you need is a screenshot of the outfit you'd like to try — whether from Pinterest, Instagram, or another site — along with a full-body photo of yourself taken in bright lighting, featuring a natural pose and no headwear. After uploading both, you can have Doppl generate a still image of you wearing the selected clothing. The generation process takes a moment, but once it's complete, you can tap the animation icon to add a random movement. This might show you flashing a peace sign, smiling and waving at the camera, or striking another casual pose.

For my test, I uploaded a straightforward photo of myself in a T-shirt, shorts, and socks. In my first try-on, I picked one of Google's sample outfits. The app reproduced the white and blue striped shirt fairly well, but it swapped the intended skinny jeans for red shorts and wrapped what should have been denim fabric around my calves, creating a strange leg warmer effect. Another outfit I tried, featuring distressed jeans, resulted in Doppl only generating the lower portion of the pants while elongating the shirt to end precisely where my shorts do in reality.

The results got even more unusual when I gave Doppl an outfit image showing a person from the knees up, wearing a striped button-down and long striped shorts. Rather than recreating a similar look, it made the shorts even shorter and gave me a pair of surprisingly believable, yet entirely fake, feet.

Interestingly, even when the outfits I uploaded didn't show the model’s shoes, Doppl still generated some form of footwear for each look. (Perhaps Google's AI simply decided the outfits looked better with bare feet?)

Throughout my testing, I noticed Doppl prevented me from uploading pictures of more revealing clothing found online, such as someone in a bikini. It also blocked my colleague, Marina Galperina, from uploading an image of former President Donald Trump. These safeguards likely help curb the creation of fake images of public figures or the generation of inappropriate content.

A curious pattern emerged, however, when Marina and I uploaded mirror selfies to virtually try on outfits. Instead of maintaining a close resemblance to our actual appearances, Doppl made both of our digital avatars noticeably thinner, to the point where we looked somewhat like bobblehead figures. This issue did not occur when I used other full-body photos taken by another person.

Google has offered a virtual try-on tool for a few years, but earlier this year it enhanced the feature, allowing users to upload their own photos and use AI to visualize themselves in tops, dresses, skirts, or pants found via Google Search. Doppl represents an even greater advancement, enabling you to try on a wider variety of clothing from different online sources and even generate a video of the result. If Google can iron out some of the tool's current oddities, I can see it becoming a genuinely useful way to picture yourself in any outfit you spot online.

Doppl is now available in the U.S. via its app on both Android and iOS.