Google DeepMind unveils BardVision XL for spatial design and AR prototyping

AI · 6 min read

BardVision XL accepts architectural plans, 3D assets, and simple sketches to generate spatial UI placements, interaction flows, and annotated anchor points for AR devices. Outputs include Unity-ready scene graphs, interaction scripts, and accessibility annotations for spatial audio and haptics.

Google is packaging the model with a lightweight authoring tool that runs in a browser and exports to ARCore, ARKit, and Unity projects. The tool focuses on realistic occlusion, scale suggestions, and collision-aware placements to reduce the iteration loop between design and engineering.

DeepMind has also published a set of research papers on evaluation metrics for spatial usability and introduced a developer preview program for partners in retail, healthcare, and industrial training. The BardVision API will be available via Google Cloud with enterprise SLAs.