DCGAN is initialized with random weights, so a random code plugged in the network would crank out a totally random picture. Even so, as you might imagine, the network has millions of parameters that we can tweak, as well as the intention is to locate a placing of these parameters that makes samples created from random codes appear like the instruct