Embodied, tangible interaction can make complex AI processes intuitive and engaging without technical jargon—by letting people physically manipulate outputs, they experientially understand how diffusion models gradually refine noise into coherent images.
Diffusion TV is an interactive art installation that lets people physically experience how diffusion models work by manipulating a modified CRT TV's antenna to control image clarity. As visitors tune the antenna and knob, they directly engage with the denoising process—the core mechanism behind AI image generation—while exploring AI-generated animals across past, present, and future timelines.