New AI Deepfakes Technology: Recycle-GAN Unsupervised Video Retargeting

Another milestone achieved in Deepfakes sciences. As this Deepfakes method will not only transform facial expressions but will also transform textures, weathers and much more. Researchers at Carnegie Mellon University, USA, have developed a new method for automatically transforming the content of one video into another by transferring facial expressions from a human character to a cartoon character or even allowing Barack Obama to speak in the same way and with the same gestures as Dr. Martin Luther King.

Deepfakes, New AI Deepfakes Technology: Recycle-GAN Unsupervised Video Retargeting, Optocrypto

One of the peculiarities and advantages of this method is that it does not require human intervention, which gives it the ability to quickly transform an enormous number of video samples. Similarly, producers of series, films or documentaries can have a big ally if this technology is brought to market for its massive use.

The researchers have also stated that the tool could be used to convert black-and-white films to a color format and to create new content for virtual reality environments. Aayush Bansal, a researcher at the Robotics Institute of Carnegie Mellon, said in this spirit:

“I think there are many interesting projects to describe. It is a tool for the artist that gives them an initial model which they can work on.”

To create this system, the preferred tool is Artificial Intelligence (AI), with one algorithm that recognizes characteristics that match the style of a video or image, and another that learns to create videos and images that match a particular style.

Even though this new method raises concerns about creating deepfakes and spreading the false news on the Internet, Bansal says that the main motivation for developing this method is the creation of movies. But the analyst acknowledges that his technology can be used to create deepfakes and induce a person to say things they didn’t say.

One of the most important things about this technology is the precision with which it converts videos, because it is not limited to changing simple expressions, but is also limited to other elements within the video, such as flowers, clouds, or designing the video environment in accordance with the interests of the audiovisual content producer.

 

Source