How to control the output of a GAN network?

Question

Neural networks

Greetings. Understanding GAN networks. For example, there is a network that draws numbers from mnist. As far as I understand, we apply random noise to the generator input in order to provide a variety of output images. Once the network is trained, we can generate random numbers similar to the mnist dataset. But how can I know what the network wanted to represent (label, the number corresponding to the picture) or, even better, how can I tell the network that I need different versions of a particular number?
I want to use GAN to create my own dataset to train another network, but if the output is random and without labels, then there is obviously no point in this.

Edit: Using ConditionalGAN, you can set a label for the network, and the output will be like this. But the question remains, if I want to generate numbers from several digits, I can’t allocate each number to a separate class - for 6-digit digits it will be 999999 classes. How to be?