This course invites hackers from creative and technical disciplines who share an interest in experimental use of AI. It is for anyone interested in understanding the inner mechanics of diffusion models and in developing strategies to disrupt and creatively misuse these systems. As a hands-on dive into state-of-the-art image and video generation models at the algorithmic and architectural levels, this course guides participants in developing custom modules for ComfyUI, such as samplers, noise schedulers, tensor operations, and noise injection nodes, and publishing them in its public library. We'll then use these tools to intervene in the diffusion process and create our own glitch aesthetic. We will shift from 'prompting the model' to actively reprogramming or breaking it.

Open to students from all faculties, the course is designed to bring together participants from artistic fields such as art, design, and architecture, and from technical backgrounds including computer science and HCI. Collaboration between these disciplines will take place through small group projects.

Further information and registration: https://www.uni-weimar.de/qisserver/rds?state=verpublish&publishContainer=lectureContainer&publishid=68296