DCGAN is initialized with random weights, so a random code plugged into the network would generate a very random graphic. Having said that, when you may think, the network has many parameters that we could tweak, plus the target is to locate a placing of these parameters that makes samples generated from random codes look like the training info.Wil