How Deepfakes Created: Rashmika today, you tomorrow! How are Deepfake videos made? – Bengali News | How Deepfake Videos Are Created Know Details As Rashmika Mandanna Incident Sparks Controversy

Deepfake heavy danger!

He is not, but there is! Actress Rashmika Mandana’s viral video has shown how much danger can be put in people’s faces by putting one’s face on another’s face. The technology is called AI Deppfakes. This dangerous method knows all the ways to imitate the exact person, including the appearance, voice of the voice. Union Minister Rajiv Chandrasekhar himself has said of deepfake, ‘a more harmful and dangerous form of misinformation.’ As Rashmika Mandana’s video went viral, stars like Amitabh Bachchan came forward in support of the actress. Big B appealed to the authorities to take strict action on the matter. The Union Ministry of Electronics and IT then advised platforms like Facebook, Instagram and YouTube to remove such misleading videos created using AI Deepfake within 24 hours.

What is deepfake?

Deepfake is a program based on deep learning, a type of artificial intelligence. This method basically creates fake images, audio-videos. Anyone can create a fake photo or video of a politician, celebrity or popular athlete with the help of deepfake technology. For a bit of fun, one can also sing or dance to their favorite or unloved star to their own tune. Not only video, DeepFake can also manipulate audio extensively.

Deepfake is actually synthetic media, which is created artificially. In this process, with the help of artificial intelligence, the face and voice of the character of the original image or video is copied. Technology has such positive uses. This technology can be used to create visual effects in movies, augmented reality avatars, multiple social media tasks, and even in education. Unfortunately, since DeepFake came to light, it has had a lot of negative uses. It has been used more for political misinformation, revenge porn, lewd videos about celebrities, to deceive the common man.

How deepfakes are made

DeepFake’s algorithm finds the similarity between two faces (the original video face and the face that will be placed there). Roughly tries to match their common features. Deepfake videos don’t emphasize close-up shots of characters. That is, such videos are found, where there are almost no close-up shots.

There are different sets of decoders, which are trained for both faces. Now the creator just needs to swap the faces from his respective decoder to create a deepfake. To make things easier, a compressed image of person X is fed to the decoder, which has actually been tested or already trained on person Y. The decoder then reconstructs X’s face based on Y’s expression.

The same process is used frame by frame to make that look more believable. That’s why you will see, the length of Deepfake video is not very big. Because, frame by frame editing of large videos is impossible. Deepfakes are also created using a method known as Generative Adversarial Networks or Gan. In that case also two AI algorithms are added to get the perfect result.