Samsung's Deepfake could make dark monk Rasputin sing like a real thing

Researchers from Samsung's AI research center in Cambridge, United Kingdom, together with computer science experts at the Royal College of London have teamed up on a project to create an anti-network Generative adversarial network (GAN), can create effects and synchronize to the point of perfect movement of the face of 2D images with sound clips including human voice.

In addition to synchronizing the movements of the lips and mouth, this artificial intelligence model can also adjust the movements of the eyebrows as well as blink to make the faces it creates seem more natural. However, the most important task will still be to synchronize the lips and mouth with the sound emitted. Today, this is usually done during post-editing, or through the use of computer graphics.

1. AI models use aerial and ground data to navigate areas that are difficult to observe

Samsung AI Center

Researchers believe that this model can be used to automatically create talking faces for cartoon characters in the most natural way, improving visibility in bandwidth video calls. Low, also provides the ability to synchronize oral or better voiceover for foreign movies.

Among the examples of this study shared on YouTube, the most impressive is the video depicting the infamous Russian mystic Rasputin (1869 - 1919), "expressing" the female song Halo. Beyoncé. Some other videos are also notable as the two late rappers 2Pac and Biggie are singing their work again, or the great scientist Albert Einstein is reading a sentence about the common language of science.

1. Nvidia's STEAL AI offers the ability to support inference for better computer vision models

Dark monk Rasputin is playing Halo

In addition, there are many other interesting example videos, you can refer to: https://sites.google.com/view/facial-animation/home

News of the study came just a month after another Samsung AI research center in Moscow introduced a project to create 2D animations using artificial intelligence without using models. 3D. It is known that this technology can be used to create digital avatars or more convincing deepfake images.

The use of GAN networks to fake digital media such as images and videos is emerging as a remarkable trend in the AI ​​field in the world over the past few years. Many sophisticated GAN models have been introduced in the framework of intensive artificial intelligence conferences such as ICML, CVPR, and many other remarkable events since the beginning of the year.

1. 5 genius minds in the AI ​​field join hands to create awesome robots

Rasputin and like singer Beyoncé

Among them, some notable models with high potential applications, such as Nvidia's GauGAN, can turn MS Paint-style doodles into artistic masterpieces. Or CollaGEN, a method created by Korean researchers to replace the missing data in each image.

Although the potential to contribute in many different areas of deepfake tools is huge, the potential risks that they are also causing problems for scientists.

In fact, this technology is often used to fake celebrities (actors, singers, models .) in false videos or worse in porn movies. As can be seen, this technology has created a strong shock because many people believe that it is true. This not only affects the honor and dignity of the rogue but also causes public opinion to become tumultuous and controversial or perhaps even worse.

1. Japan's artificial intelligence has created ultra-realistic virtual fashion models

Deepfake contains many potential risks

Perhaps in the future, scientists and state management agencies need to sit together to find conventions for using this wonderful technology in a proper and safe way for the community.