Take a video and replace the face in it with a face of your choice. You only need one image of the desired face. No dataset, no training.