Google has launched Doppl, an experimental mobile app that uses AI to simulate clothes virtually, in the US on iOS and Android platforms. Users can dress their digital avatars in a variety of outfits by uploading just one full-body photo, and the app can also be created as a video.
The app works by combining photos or screenshots of the desired outfit with a full-body image of the user, and AI generates a virtual outfit image. For example, you can use photos of clothes taken at a second-hand store or images of styles saved from social media. In addition to simple still images, you can also create videos with animated movements such as hand gestures or smiling, allowing for a realistic style check.
Doppl is an extension of the virtual fitting technology first introduced by Google Shopping. While the previous technology focused on dressing models of various body types, this time it focuses on the user’s own avatar.
However, since it is still in the experimental stage, the exact fit or the detailed expression of clothing may be somewhat inaccurate, and there have been reports of cases where long pants are incorrectly expressed when taking a picture wearing shorts, or socks are replaced with virtual feet generated by AI. In addition, uploads of images with a lot of exposure, such as bikinis, or photos of certain people (e.g. former President Trump) are restricted.
Google plans to use Doppl to help users explore new styles while also gathering data to develop AI-based fashion and shopping technologies in the future. The app is currently only available for download in the United States, and there is no date for its release in other regions.
