What problem does it solve? Building AR experiences on Spectacles that react to real-world geometry requires per-pixel depth data, but the DepthModule is an experimental API with non-obvious setup requirements, frame synchronization challenges, and no depth streaming in Lens Studio Preview, making integration error-prone. ## Core Features & Use Cases - Raw Depth Frame Access: Create a DepthFrameSession, receive per-pixel depth in centimeters, read camera intrinsics, and unproject pixels into world-space 3D points. - Depth Cache Pattern: Snapshot a synchronized depth+color frame pair, send the color image to an AI vision model, and convert returned pixel coordinates into world positions for placing AR labels on real objects. - Depth Texture Visualization: Render the live depth feed as a world-anchored point cloud using a custom instanced material for debugging and verification. - Use Case: A user pinches to capture a scene, the color frame goes to Gemini for object detection, and each returned pixel coordinate is converted via the cached depth frame into a 3D anchor for an AR label. ## Quick Start Ask the agent to set up a DepthModule session on Spectacles that reads depth frames and prints the world-space 3D position of the center pixel.
