ÎÞÂëÊÓƵ

Editing Reality explores how generative AI can become part of an embodied, situated creative process in mixed reality. Instead of treating generation as a one-shot command on a screen, the system allows users to work inside the room they are modifying: they can speak or type prompts, sketch 3D forms, select objects, preview changes, and revise generated results directly in context.

The system supports in-situ generation, reconstruction, erasing, retexturing, sketch-based modification, iteration, and multi-user collaboration. It is designed for early-stage spatial ideation and situated prototyping, where designers can rapidly explore furniture, material changes, object transformations, and speculative modifications at 1:1 scale within the target environment.

Through iterative system development, a formative workshop, and expert review, the project frames generative MR authoring as a negotiated, spatial, and temporal process. The work highlights how users negotiate intent with the system, accumulate design history in space, evaluate results through bodily movement, experience waiting at the site of change, and gradually learn the AI system as a spatial actor.