Samsung’s Deepfake AI Is Capable Of Creating A Convincing Video Out Of A Single Frame

Now Reading
Samsung’s Deepfake AI Is Capable Of Creating A Convincing Video Out Of A Single Frame

Browse This Page
Share via

Artificial Intelligence (AI) is expected to be the next big thing in technology in the near future. The idea of human-made objects able to exist as sentient beings capable of making decisions is both optimistic and terrifying at the same time. In a development bordering between the aforementioned emotions; Researchers at the Samsung AI Centre in Moscow, Russia have developed a Deepfake algorithm that can create a video out of a single image. These “Living portraits” as they are called have been tested on some famous historical images and the results are well, surprisingly great.

Samsung Deepfake AI

As mentioned previously, Samsung has developed and had (kind of) successfully tested the algorithm. In a video shared on YouTube, the AI can be seen in action. The researchers refer to this process as few/one-shot learning. In this process, a model can be trained by using only an image to chart out a convincing animated portrait shot. While this works with a single image, using more images, up to 32 improves the results even more. This process of superimposing existing images and videos onto source images or videos using a machine learning technique is generally known as Deepfakes. This has been an avid problem on the internet today which is credited to the rise of fake news.

Samsung Deepfake AI

Also read: LEGO’s New Team-Battle RPG Announced By Gameloft For iOS And Android

These realistically talking head models are generated using a convolutional neural network. By training the algorithm on a large dataset of talking head videos with a variety of appearances. The Samsung researchers have been reported to have used more than 7,000 publically available images from YouTube videos. This technique generated Deepfakes of historical personalities such as famous scientist Albert Einstein, Painter Salvador Dali and the mysterious woman from the Mona Lisa portrait of Leonardo Da Vinci. And while there are a few anomalies that give away the obviously fake nature of the video, it is an example of how far the technology can go. In the meanwhile, we will just leave this GIF of expressive Mona Lisa for you to ponder.

What's your reaction?
I Want This
About The Author
Avatar photo
Yashonil Gupta
Easily Distracted By Meta Objects, Movie Nerd And Theory Inventor. Yash Covers All Tech at iGyaan