// new media class

I explored two methods of developing images: the latent diffusion framework Stable Diffusion and my own process of making digital automatic drawings.

The diffusion framework works, in part, by taking a set of given parameters (prompt, step count, etc.) and starting to generate a discrete sequence of images. At each iteration of the sequence, the image becomes more and more refined until converging at a fixed state or reaching the final step of the sequencer. Observing this progression, one can notice how some features or areas of the image gradually settle into place, while other parts continue to shift as the framework recalculates and refines details to better align with its predicted target result.

In making my automatic drawings, I try to keep a ‘distanced influence’. The process involves many cycles of building up an image, subtracting from it heavily and then building it back up again.

To start the tandem cycle, I generated an image with Stable Diffusion and started an automatic drawing in parallel. Next, I worked fragments of my drawing into an early step from the sequencer and initiated a new iteration from that point, seeing how the changes to the given step would alter the outcome of the image generation process. I repeated this process several times, each round splicing my drawings into a later step in the iteration.