How deepfake app Zao became most-downloaded free app in China, and what's next

Zao’s deepfake technology lets users modulate the voices of celebrities as well as stitch their faces onto an actor’s body in a scene, turning the likes of Brad Pitt into a middle-aged banker from Shanghai or a sorghum farmer from Inner Mongolia.

Every year, Google receives billions of search queries pertaining to what clothes celebrities wore on the ramp, in movie scenes, on date nights, or when taking their pets out for a stroll. Many such searches link back to the websites of textile brands whose wares are reproductions of the tailor-made outfits that line celebrities’ wardrobes. For online vendors selling consumer goods such as shoes, apparel, or spectacles, the next frontier is the virtual showroom, wherein users will be able to sample their wares for trim and finish, on mobile devices, rather than in trial rooms.

With the emergence of deepfake technology, superimposing one’s face onto another’s body, especially in video clips, has become much easier. But this comes at considerable cost to privacy. A Chinese app, Zao, has gained considerable popularity in a short time span by letting users morph their faces over that of their favourite actors, placing themselves in short clips from movies and TV shows. This is analogous to similar experiments with audio, such as Dubsmash.


However, the app – intended to drive lighthearted banter on social media – places the tools of creating deepfake videos in the pockets of millions. Zao is available to download on Android and iOS for users with a Chinese mobile number. The technology lets people modulate the voices of celebrities as well as stitch their faces onto an actor’s body in a scene, turning the likes of Brad Pitt into a middle-aged banker from Shanghai or a sorghum farmer from Inner Mongolia.

But aside from creating seemingly frivolous and innocuous content, deepfakes pose an existential threat to mainstream media, as the technology can be leveraged to spread misinformation by putting words into the mouths of influential people who wield significant power on world affairs. It also undermines the privacy of individuals, as their social media accounts might have a trove of personal photos, which are rife for harvesting by third parties with malicious intent.

Zao was developed by a subsidiary of Chinese software company Momo, and its popular has soared since debuting last week. The app is now the most downloaded free app in China, as per the latest rankings released by analytics company App Annie. Zao lets users select a scene and then upload a photo of themselves, and in a few seconds, the algorithm renders the media files to stitch together a video clip adding the faces of users to that of actors from popular movies and TV shows, both Chinese and foreign.

Momo is not the only developer which is helping deepfake technology go mainstream. A rudimental search on Google throws up different tools that offer the basic software to create deepfake videos, be it a mash up of Angela Merkel and Queen Elizabeth, Elon Musk bragging that “he is in advanced talks to buy Volkswagen,” or Lionel Messi and Cristiano Ronaldo coming to blows on the sidelines of FIFA’s annual awards night. While some such content might seem primitive, the proliferation of apps like Zao have increased the turnaround time of churning out such videos, raising concerns over how they could be used to peddle fake news online.

Privacy critics argue that users of such apps may be inadvertently feeding a repository their personal photos to the databases of companies with questionable motives. They can be used to stoke controversy, be sold to advertisers for targeted marketing, or even for creating pornographic videos. The mass production of such content also raises questions over who owns a person’s image – especially since celebrities command upwards of eight-figure sums for image rights, and clips from movies are copyrighted by production houses. The Russian face-swapping app FaceApp also faced similar scrutiny in July.

FaceApp spawned a viral trend, #OldAgeChallenge, where people ran selfies captured in real time through the app’s filter to generate images portraying how they’d look in their dotage. The hashtag was popular across social media platforms like Facebook, Twitter and Instagram. The app’s popularity in the United States, and its ties to the Russian state drew criticism from many quarters, even prompting an FBI investigation. Wireless Labs, the creator of FaceApp clarified that it allowed users to delete their photos and that - by default - images were deleted with 48 hours of their upload date.

China’s Zao has had to fend off similar concerns about its terms of service. The company initially said that it reserved the right to use images generated using the app, but later watered down its policy, saying that it would not use content created by users without their permission. Zao further added that upon deleting the app, a user’s personal images and all mashups created using the software would be removed from the company’s servers. However, it remains to be seen whether such information can be permanently scrubbed clean from the internet.