The generative AI model introduced by NVIDIA in 2018 that became famous for realistic synthetic human faces was StyleGAN.
StyleGAN was developed by researchers at NVIDIA and presented in a 2018 paper by Tero Karras, Samuli Laine, and Timo Aila. It adapted the generative adversarial network approach by giving the generator more control over image features at different scales, such as pose, hairstyle, facial structure, and fine details.
The system became widely known because it could produce highly realistic portraits of people who do not exist. Its results helped demonstrate how generative models could synthesize convincing faces, while also raising concerns about deepfakes, identity misuse, and the difficulty of distinguishing generated images from photographs.
StyleGAN is sometimes confused with FaceNet or DeepFace, which are primarily associated with face recognition rather than image generation. NVIDIA later released improved versions, including StyleGAN2 and StyleGAN3.