Artificial intelligence for image generation has made enormous progress in recent years. Systems such as Midjourney, ChatGPT and Stable Diffusion generate impressive images from simple text descriptions. But how does this process work in detail? What happens in the ‘black box’ between text input and the finished image? 

In our presentation, we will open this black box and make the hidden processes visible. We will present the results of our design research, which deals with the visualisation of the underlying mechanisms of image-generating diffusion models. In particular, we focus on the ‘latent space’ - the abstract construct that controls image generation. We show how design approaches can make complex technical processes accessible and how design acts as a bridge between technology and human understanding.