There are a number of image-based methods we have conveniently broken down into categories:
Full face generation - GAN models like stylegan2 and stylegan3 and Diffusion models like Stable Diffusion, Midjourney.
Partial face generation - Face Swapping (given a source face and a target image with another face, the source face is swapped into the target face). Face Swapping also uses GANs and generative-based deep learning to further improve the face blending. There is also a technique called in-painting which leaves a hole in the head and the deep learning method fills it in.