// 2026.05.21

Giving Meta Ray-Ban Glasses Precise Location: MultiSet VPS

10:00 JST

A through-line of this series is that AI glasses get dramatically more useful once they know precisely where they are. As MultiSet AI’s Nikhil Sawlani frames it: precise location context unlocks turn-by-turn audio navigation, location-aware games, and “finding your Uber pickup spot” — the right information exactly when and where you need it.

MultiSet’s Visual Positioning System (VPS) delivers that location awareness to Meta Ray-Ban Smart Glasses, and there’s a working iOS sample to learn from:

▸ github.com/MultiSet-AI/wearable-vps-samples (iOS)

What the sample does

It pairs with Meta Ray-Ban glasses and combines the MultiSet VPS API with the Meta Wearables Device Access Toolkit (the DAT SDK from our Meta Ray-Ban Display post):

  • Localization — capture a frame from the glasses camera, send it with camera intrinsics, get back a 6-DOF pose (position + orientation) and a confidence score within a mapped space.
  • Turn-by-turn navigation — waypoint graph + A* pathfinding, with directional audio played through the glasses’ speakers to Points of Interest.
  • Multiplayer — streams the wearer’s localized pose (~20 Hz over Multipeer Connectivity) to a host running the MultiSet iOS SDK, for shared-space AR.
  • Low latency — session pre-warming, on-device downscaling, and a video-frame fast path for re-localization.

Try it

Prerequisites: iOS 17+, Xcode 15+, Meta Ray-Ban Smart Glasses paired in the Meta AI app with Developer Mode on, the Meta Wearables DAT SDK 0.6.0 (resolved via Swift Package Manager), and MultiSet VPS API credentials from the MultiSet Developer Portal. You also need a mapped environment — maps and navigation data are produced with the MultiSet Unity SDK (NavMeshExport scene → {mapCode}_navigation_data.json).

git clone https://github.com/MultiSet-AI/wearable-vps-samples.git
cd wearable-vps-samples/iOS/MultisetWearable
open MultisetWearable.xcodeproj

Add your Client ID/Secret (xcconfig + Info.plist is the recommended, secret-safe path), set your Map Code in the in-app Settings, pair the glasses, and run the Navigation Demo.

Note: unlike the open-source toolkits earlier in this series, the MultiSet SDK is under a proprietary license and requires API credentials. It’s free to try via the developer portal; check terms for production use.

Why it matters for the hackathon

Localization is the missing primitive for genuinely spatial glasses apps. With VPS you can build a venue navigator, a location-anchored game, or a shared-space multiplayer demo on hardware people actually wear — without inventing your own tracking. It pairs naturally with the Meta Ray-Ban Web App approach from earlier in the series.

Thanks to Nikhil Sawlani (MultiSet AI) for surfacing this. Questions? Reach us via the Contact page.

// BACK TO NEWS