Speak a scene into VR.
Active Developmentto Present
Build and revisit a VR scene by describing objects, pointing, and moving them by hand.
This two-person research project explores reconstructing remembered places in VR. Speak an instruction and point to a location, then move, rotate, or resize the result by hand. Objects come from a library; missing ones can go through image generation and TRELLIS to become meshes. Worlds can be saved and reopened.
Current prototype
The lantern images show a completed image-to-mesh test. The voice panel is from September 14, before the latest hand-menu revisions. Current work is hand tracking and headset interaction. The prototype is unreleased, and no user study is complete.
My contribution and AI use
A two-person research project with AI-assisted development. My recent work is the implementation and integration of the VR prototype; Maurice focuses on the supporting literature. The current planner uses OpenAI. The screenshots record the September 14 development build and generation test, before later hand-menu revisions.