At fake apps became popular during 2020, turning any smartphone into a real animation studio, capable of creating the famous “deepfake” videos. These applications may require a paid subscription, but most of them offer free versions with which the user can easily put the image of anyone, doing and saying things he has never done or said.
And so, what many feared, happened. The trivialization of a tool that can be used as a weapon of defamation, aggression and manipulation, with very serious consequences. Both for those who see their image manipulated, and for those who believe and act on the basis of these false images.
Fake apps: how they work
The name deepfake emerged from a combination of deep learning – designation given to the deep learning process performed by Artificial Intelligence software – and the word fake (false).
In fake apps, the deep learning defines the way algorithms recognize patterns from real images, photographs or videos, about a particular person. The algorithm learns “what that person is like” and will reproduce their image in a new context.
As new data is introduced into the application, for example something we want the person to say or a movement we want them to make, the algorithm executes it, constantly comparing the graphic result with the original images, correcting and optimizing the patterns found.
This reproduction will be all the more perfect the greater the number of images that the algorithm had access to about that person. He will then be able to learn how she moves, what facial expressions she has, what movement she makes, and even what her voice is like, and then reproduce them in the new context.
Fake apps: the most famous
While the primitive software used in animation cinema produced these effects in a slow and exhaustive way, to obtain a perfect final result, the new fake apps do it in a cruder way to be faster and simpler. But they do it better and better.
Here are some examples of the most popular applications.
This application was the one that most popularized the creation and sharing of deepfakes on the social networks.
Despite being only available for the Chinese market, it has become very popular in this country and the videos produced by its users have invaded the internet, especially TikTok.
Very easy to use and very quick to synchronize facial expressions, the application allows the user to place the image of their face or anyone else’s, in any type of video, movie or TV show, creating very realistic results.
In its short and successful period of life, this application had only one setback, which acted as a warning for users of this type of software. MoMo, the company responsible for Zao, initially owned the users’ images, being able to use them freely, but was forced to change its policy. Currently, the company undertakes not to use these photos, deleting them from the server if the user does so in its application. But still, be warned.
The database of zao it may already be huge and it looks like it will grow when it becomes available to users outside of China.
Reface: the fastest face swap in the west
This application uses the technology faceswap to swap the user’s face with someone else’s in a video or animated gif. The user just needs to take a selfie and their face is mapped onto another face in a very realistic way.
Reface is one of the most downloaded apps on the American App and Google Store. Launched in January 2020, in September it had more than 42 million downloads from the App Store.
Avatarify: the most popular app for iPhone
It’s never been so easy to put a celebrity or someone else’s photo saying and doing what the user is saying and doing in real time. Just open the photo of the desired person and turn on your selfie camera. Then just talk and move to see the screen image repeat exactly what you’re doing, as if it were a puppet at your fingertips.
Videos made in the free version show a watermark, but if the user subscribes, it no longer appears. It’s not long before Avatarify become available for Android as well.
Wombo: It’s never been so easy to be a singer
This application is specialized in lip synchronization, which allows the user to transform a photo of himself or someone else, into a moving video, where they appear to sing.
In just the first two weeks of use, this app, available for Android and iPhone, generated more than 100 million videos.
THE Wombo it offers several types of music by default so that the user can select and apply them to the chosen photo.
My Heritage: resurrects old photos
The My Heritage site allowed users to discover family history, develop their family tree, find new relatives and explore hundreds of historical records. But now it also offers a new fake app feature, Deep Nostalgia, which allows users to bring their older family photos to life.
The application has a 14-day free trial period, but its users are growing by leaps and bounds. According to official sources, more than 82 million animations have been made with the Deep Nostalgia and the number continues to grow.
Fake apps: deepfake needs rules
The technology for digital image manipulation has been available to the general public for several decades, as is the case with the famous software Photoshop. But the ease with which these new applications do it with video has changed the rules of the game, turning it into a dangerous weapon for unethical users.
Fake news, harassment or even the creation of fake pornography are the most reported problems and, so far, women have been the biggest victims.
the first big boom of use of deepfake was made using the image of actresses, singers and famous female personalities that were mapped and placed in adult video scenes. The call Revenge Porn (revenge porn), which has also affected women in the political scene and even non-celebrity women, is considered one of the most frequent and harmful uses of fake apps. (source: deeptrace)
But we are only at the beginning of the massification of this technology. Both governments and ethicists are already showing great concern for what is yet to come. They fear that it could contribute to an increasing discredit on democratic institutions and appeal to the companies that develop these applications to take measures to define rules and uses that allow them to circumvent these problems.