This paper proposes a method to create objects for use in the metaverse using generative adversarial networks (GANs) coupled with an autoencoder. Specifically: 1) A GAN model called DCGAN is used to generate new images of human faces from a dataset of over 5000 images. 2) The lower quality images produced by the GAN are then upscaled using an autoencoder model to improve image quality. 3) The higher quality generated images can then be used as virtual objects that are more connected to the real world in augmented and virtual reality applications of the metaverse.