Hands-free AI for blind and low-vision people

The AI that sees, remembers
and acts for you

Talk to the bracelet on your wrist. The phone on your chest watches the way ahead. GuideEye warns you, remembers for you, and pays only after your spoken yes.

Free and open source · Android, macOS, Windows, Linux and the web

In your ear

Four things you can just ask for.

No screen to aim, no buttons to find. Pick one and hear how the conversation goes.

Try an example

Ready · listening

“What’s in front of me?”

Chair ahead, slightly left. The doorway is clear beyond it.

Detected on device

Chair · ahead, left

Processed locally. No video stored.

What it does

See. Remember. Act.

Today’s vision apps make you aim a phone and tap buttons — then forget everything the moment you close them. GuideEye works the other way round.

See

Live hazard warnings spoken as you move: “Chair ahead, left.” Detection runs locally, so it stays fast and private.

Remember

Ask later: “Where did I leave my keys?” GuideEye links what it sees into a graph — keys on the counter, counter next to the kettle — so it can answer from the connections, not just the words.

Act

Say “Buy me a coffee.” It reads the item and price aloud, then pays only after you say yes.

Hands-free by design

The bracelet listens and speaks, the phone stays clipped to your chest. Your hands stay free for your cane, your dog, your day.

Bridge Coming soon

A bridge between deaf and blind people: sign language voiced aloud, and speech shown as large captions.

Made for AI wearables

GuideEye is an accessibility layer that plugs into AI wearables — starting with the Vexo bracelet.

How it works

The bracelet is your voice. The phone is your eyes.

Safety-critical detection runs locally and privately. Reasoning and natural voice use the cloud.

  1. 1

    You talk

    Speak to the Vexo bracelet on your wrist. A chest-clipped phone camera watches the way ahead.

  2. 2

    It understands

    Hazards are spotted on-device in real time. An AI agent handles questions, graph memory and requests.

  3. 3

    It answers — or asks

    You hear a clear spoken reply. For anything that matters, like paying, it waits for your “yes”.

Live demo

Try it. This one is real.

Your camera, your voice, the real GuideEye service. Turn the camera on, then ask a question out loud or type it.

Checking the GuideEye service…

Vexo band not connected

Camera is off

1×

Microphone off

  1. Hi, I’m GuideEye. The microphone is on — just ask what’s in front of you, where you are, or to pay for something.

To try paying by code, show the camera this demo payment code on another screen and say “Guide, pay this”.

Camera frames go to the GuideEye service on your own computer. When you ask a question, that single frame is also sent to Google Gemini for the answer. Directions send your location to OpenStreetMap services. Speech is recognised by your browser. Nothing is recorded. Payments are test-mode only.

Built for trust

Nothing happens behind your back.

An assistant you can’t see has to earn your confidence. GuideEye is designed around three promises.

  • Private by default

    No raw video or audio is stored.

  • You confirm

    Money never moves without a spoken yes.

  • Honest about doubt

    When it isn’t sure, it says so instead of guessing.

Please note: GuideEye is a prototype. It does not replace a white cane, guide dog, orientation training or emergency services.

Get the app

GuideEye, built for your device.

We’ll detect this device and put the right build first. You can always choose another platform.

Checking your device…

Finding the right GuideEye build

This takes less than a second.

Open web app

GuideEye 0.1.0 · Early hackathon prototype. Desktop builds are unsigned, so your operating system may ask you to confirm before opening.