Did you listen to Kim Jong-un's speech about democracy, or admire Salvador Dali, who talked about his death (why doesn't he believe in it)? Or were you shocked by the video with Zelensky about "surrender"? These videos have one thing in common - they are fake. Deepfake is a technique for creating and publishing false information in the form of video, audio and photos. It is becoming easier and easier to come across such materials, so we have prepared for you instructions on creating, recognizing and searching for deepfake videos on the Internet.
Deepfake – what is it?
Deepfake is essentially a photo, video, or audio material that is not real, fake, or fake. Deepfake (a portmanteau of the words deep learning and fake) uses artificial intelligence (AI) to create a deep-learning look uruguay number data that mimics one person in video and other digital media. It is created by algorithms based on real-world voice, sound, video, or photo samples that are stitched together. The algorithm takes selected information from many different places and then combines it into one material. The result is something new (fake) that is based on a combination of different real-world data.
But that's not all. Deepfakes can also "paint" a photo from scratch, for example, creating the faces of characters who never existed.
What is the purpose of such actions? Among other things, Deepfake is used to spread incorrect information online (for example, about political figures or events), theft (for example, forging voice commands), and the creation of pornography (celebrity movies - deepfake porn). It can also be used for educational and entertainment purposes.
Deepfake – how does it work?
In general, there are two ways to create deepfake material. The first method uses two algorithms. First, the first (encoder) looks for common features between two images (those that need to be merged into one), then the second (decoder) can transfer them to the newly created image. For example, if we want to create a video in which an actor moves like us, the encoder takes the movement from a video where we dance, and the decoder will replace our face with the actor's face and make him reproduce our movements. The key here is the selected actions that are copied to the new image (for example, movements, gestures, facial expressions).