Skip to Main Content

HeyNavi

Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

HeyNavi — Demo video

HeyNavi is a voice-first iPhone companion that helps blind and low-vision people notice what is around them while they walk. Most image-description apps answer one photo when asked. HeyNavi keeps watching. Every few seconds, Muse Spark (muse-spark-1.3) reads a camera frame together with a short scene memory and picks what matters for walking: things on the floor, steps and level changes, low-hanging objects, people ahead. SAM 3.1 then checks each item on the same frame and outlines it. An obstacle is announced only after SAM confirms it, and anything unconfirmed is hedged ("There may be..."). The app decides what is worth saying, skips repeats and otherwise stays quiet, and you can cut it off at any time by voice or touch. You can also just ask in plain English, like "Where is the exit?" or "Which room is this?". Muse Spark answers from a fresh photo, and SAM outlines what you asked about. A live decision panel shows each step of the loop. Measured on an iPhone 16e in our own test runs: one look takes about 4.8 s round trip (median), SAM confirmed 83% of the items Spark proposed, and a stop command silences the voice in about 17 ms once the app receives it. HeyNavi never says a path is safe, and it does not replace a cane or a guide dog. Open source (MIT).

Team