Dreamscape explored the intersection of two technological trends: the accessibility of mixed reality experiences using pass-through displays, and the rapid progress in 3D and 4D generative AI. Models such as Google’s CAT3D or NVIDIA’s LATTE3D show how the quality of generated output is improving while generation time and computational cost is rapidly approaching zero. What do design processes look like when 3D assets can be generated instantly and with natural language? What new design tools might be required to take advantage of these new capabilities? Mixed reality environments might be uniquely well-suited to these design tools, since there is no requirement for keyboard or mouse input when assets can be generated directly from spoken natural language and manipulated by hand. The Dreamscape workshop was interested in establishing mixed reality environments that are inherently collaborative in order to explore collective creation of new forms of architecture using collections of generated assets. Participants in the workshop were introduced to tools for working with generative AI within mixed reality environments using the HoloLens 2 and Meta Quest 3, before prototyping natural user interfaces for interacting with these APIs and working together on several small sculptural prototypes. The second day of the workshop introduced participants to wide tolerance mixed reality fabrication approaches using passthrough displays, specifically in the assembly of intricate wireframe structures using generated models as approximate guides. A sculptural object was exhibited at the main ACADIA conference, along with an immersive exhibition of the generated environments. 

Workshop Participants: Graham Boyd, Tzu-Ying Chen, Eric H. Cheung, Shawn Chiki, Po-Chun, Chou, Youqian Cui, Melissa Goldman, Richard Hua, Pearl Jishtu,
Anahita Mosayebi, Hekuran Musli, Laura Nica, Kraig Sims, Gregory, Spaw, Richard St-Jean, Simon Treml, Qingru Yang, Junling Zhuang.