The algorithm for incredibly realistic AI-powered portraits is outlined in a paper published by researchers from computer game company NVIDIA. It utilizes the Generative Adversarial Network (GAN) which was first released back in 2014 but now it operates in conjunction with artificial intelligence technology to deliver a high-quality product.
For the realistic AI-powered portraits, NVIDIA feeds the algorithm a series of photos and the technology, in turn, blends the images together. This method is referred to in the tech universe as "style transfer." The end result often boasts an entirely new person.
Aside from faces, the Generative Adversarial Network and the artificial intelligence algorithm can work together to also produce images of other objects — from fictitious interiors of buildings to cats and vehicles.