Deepfake, what is it exactly? – Geeko

“Deeplearning” is a computer method that makes it possible to create “deepfakes”. Dystopian idea or near future, these fake videos that feature personalities are more and more talked about. But in reality, how do they work?

The term deepfake is the result of a mixture of words used to create a new word: “deep learning” and “fake”. This technology uses a form of machine learning called deep learning technology. It is thus able to understand what your face looks like from different angles and uses this information to transpose a fake face on yours.

Simply put, deep learning technology is a type of machine learning that applies neural network simulation to massive data sets. Artificial intelligence (AI) effectively learns what a particular face looks like from different angles in order to transpose that face onto a target as if it were a mask. This fake face, entirely created, then becomes a mask on your face that you can control and manipulate.

These videos, sounds, texts or images are made using algorithms in the field of artificial intelligence. These are called artificial neural networks or generative adversarial networks (GANS).


Generative adversarial networks or GANs (Generative Adversarial Networks) are learning algorithms based on artificial neural networks. They are able to model and mimic any data distribution. Thus, they are used in different fields. For example, the processing of images, text, or even the creation of deepfakes.

Concretely, a GAN opposes two neural networks, one creating the fake and the other noting its efforts, teaching the synthesis engine to make better fakes. In other words, one is generator, the other is a discriminator. To synthesize the image of a fictitious person, the generator learns by iteration to synthesize a realistic face. Iteration consists in repeating an experiment until obtaining the most precise result possible. Thus, at each iteration, the discriminator learns to distinguish the synthesized face from a corpus of real faces. If the synthesized face can be distinguished from real faces, the discriminator penalizes the generator.

This technical exercise, still perfectible, is improving day by day. Tomorrow, this technology, which is becoming more democratic, could be accessible to everyone. There are already much simpler and more accessible ways to generate deepfakes.

Multiplying techniques

For the uninitiated, creating deepfakes is still a relatively complex activity today. On the other hand, there are many tools to create these deepfake videos.

There is facial reconfiguration, face change, lip reconstruction, puppet-master, face reenactment, human voice synthesis.

Face swapping, or the Faceswap technique, involves replacing a person’s face in a photo or video with another person’s face. This can be done either manually or using artificial intelligence. Lip reconstruction, or lip-sync, is a simultaneous voice and image recording technique. And, in particular, the synchronization of lip movements with the recorded sound. This process allows a person’s lip movements to be matched to words they are not speaking. Finally, puppeteering, or the puppet effect, consists of “animating a video of a person using the facial and body expressions of another person sitting in front of a camera”explains the site In short, a video A will control the movements of a video B.

Techniques that are becoming more democratic

Today the trend is towards democratization via consumer applications like FakeApp or more ephemerally DeepNude. Prior to its removal, the latter allowed the image of a fully clothed woman to be taken and undressed to create a non-consensual pornographic video.

In 2019, Stanford engineers succeeded in make video editing as easy as text editing. They had created a new algorithm that allowed video editors to edit talking head videos as if it were text editing. And this, by copying, pasting, or adding and deleting words.

Thus, the technology of “deepfakes” offers many interesting possibilities for various creative sectors. For example for dubbing or creating synthetic characters in films and video games. The technology is even being used to produce corporate training videos and train doctors. However, some fear that this technology is being used for unethical purposes. For example, to generate false political statements.

The example of ZAO

From its launch, ZAO quickly became the favorite mobile application of the Chinese. Clearly, this allows you to integrate the face of a user on the body of cult film characters, in photo and video.

For this, the user must register in the application a whole series of poses on which he must blink his eyes, move his mouth and perform different facial expressions. Next, ZAO’s technology ensures that his face matches the character’s face perfectly.

Follow Geeko on Facebook, Youtube and Instagram so you don’t miss any news, tests and tips.

We wish to give thanks to the author of this write-up for this amazing material

Deepfake, what is it exactly? – Geeko

Check out our social media profiles as well as other pages that are related to them.