Deep learning has grown significantly in the previous few years, which has enhanced computer vision applications. Deepfake is one of those recently developed apps that uses deep learning. Deepfake algorithms are capable of creating convincingly fake images and videos that are difficult for anyone to distinguish from the real thing. In this paper, we review some algorithms used to produce deepfakes, particularly for the StyleGan family. Also, we detail the difficulties, prevailing theories, and future directions in deepfake technology development. The aim of this work is to compare three variant architectures of StyleGan using the Flickr Faces HQ (FFHQ) database in order to evaluate their performance and deduce the best of them for deep false generations.