Vol. I — Issue 01 — A prototype, demonstrated at Solebury School
Faces, recognized.
A smart-glasses prototype that quietly speaks the name of the person in front of you — built for a high-school senior with face blindness.
You walk into a room. Faces turn toward you. For most people the names arrive automatically, attached to the faces like labels. For someone with prosopagnosia, the labels never appear. A familiar smile is just a smile; a parent, a teacher, a friend — all strangers, until something else gives them away.
Face Assist is a quiet companion. A button on the temple of a pair of ordinary-looking glasses. A photograph. A name, spoken only to you.
About prosopagnosia.
Prosopagnosia — face blindness — is a neurological condition in which the brain cannot reliably recognise faces, even of people one knows well. It affects an estimated two-and-a-half percent of the population.
People who have it learn to read voices, gait, hair, glasses, the order in which people arrive in a room, the small social cues that everyone else takes for granted. The compensation works, most of the time. But it is effortful, and crowded rooms — assemblies, parties, a school gymnasium — are exhausting.
Existing tools are clumsy. A phone, pointed at someone's face, is a statement. It draws attention to the very thing the wearer is trying not to draw attention to. What is needed is something quieter — something that lets the wearer keep their hands at their sides, their eyes on the conversation, their dignity intact.
“The problem isn’t that I can’t see faces. The problem is that they don’t come with names.”
- Prevalence (developmental, lifetime)
- ~2.5%
- Faces a person typically learns to recognise
- ~5,000
- Faces a person with prosopagnosia can reliably recognise by sight alone
- a few — sometimes none
Figures are approximate and drawn from clinical literature. Face Assist is not a medical device and makes no diagnostic claim.
A name, spoken only to you.
Face Assist runs on a pair of camera-equipped smart glasses. When the wearer needs a name, the device discreetly takes a photograph, runs face recognition against the wearer’s own gallery of people, and speaks the match back through a bone-conduction speaker that only the wearer can hear.
-
i.
Trigger
The wearer presses a button on the temple of the glasses, or speaks the words “who is that”. No phone. No screen.
-
ii.
Capture
The on-board camera takes a single still photograph of whatever the wearer is looking at. (A small indicator LED lights up — required by the platform, and we think rightly so.)
-
iii.
Match
The photograph is sent to a face-recognition service (Luxand Cloud) that compares it against the wearer’s private gallery of enrolled people. A match returns a name and a confidence score.
-
iv.
Speak
The name is spoken through the glasses’ bone-conduction speaker — audible to the wearer, inaudible to everyone else. The conversation continues, uninterrupted.
Prototype specification
| Hardware | Mentra Live — open-source smart glasses with camera, microphones, and bone-conduction speaker. No display. |
|---|---|
| Triggers | Temple button; voice command “who is that”. |
| Recognition | Luxand Cloud — face match against the wearer’s private gallery. |
| Feedback | Text-to-speech via the glasses’ bone-conduction speaker. |
| Gallery | Per-wearer. Each user points the app at their own Luxand account, so one wearer’s family and friends are never visible to another. |
| Latency target | Roughly one to two seconds from press to spoken name. |
If you have prosopagnosia, here’s how to use it.
Face Assist is being prepared as a MiniApp on the Mentra OS store. The flow below is what installation will look like once the demo build is promoted from Development to Published.
-
Step 1
Get the hardware.
You’ll need a pair of Mentra Live smart glasses and the MentraOS companion app on your phone.
-
Step 2
Install the MiniApp.
Open the MentraOS app, find Face Assist in the MiniApp store, and install it. The MiniApp asks for camera and microphone permissions on the glasses; it does not ask for anything on your phone.
-
Step 3
Create your own face gallery.
Make a free Luxand Cloud account. This is where the names and reference photographs of the people you want recognised will live. No-one else can see your gallery; we do not see it either.
-
Step 4
Connect the two.
Open Face Assist on your phone and paste the Luxand API token from your Luxand dashboard. From this point on, when the glasses see a face, only your gallery is searched.
Note: the in-app token-pairing screen is being built in parallel with this site. Until it ships, ask the developer directly.
-
Step 5
Enrol the people you want to recognise.
Upload one or two clear photographs of each person to your Luxand gallery, with their name. Front-facing, well-lit, no sunglasses. A dozen photos is enough to get started; you can add more over time.
-
Step 6
Wear them.
Press the temple, or say “who is that”. Listen for the name. That’s it.
Behind the build.
Face Assist is Andreas Hunt’s senior project, built with his father David. They scoped the problem, chose the platforms, tested the prototype, and curated the gallery of people it recognizes. The demo is at Solebury School, in Pennsylvania, on the 29th of May, 2026.
A high-school senior at Solebury School. The condition is his; so is the idea of doing something useful about it. He scoped what the product should feel like in real classroom and hallway situations, and he’s the one wearing it.
Andreas’s father. Specified the prototype, sourced the hardware, ran the testing sessions, curated the recognition gallery, and is hosting the demo. The shape of this project is his and Andreas’s decisions throughout.
A freelance software developer hired by David to write the code that connects the smart glasses to the face-recognition service.
Mentra provides the open-source smart-glasses platform and SDK. Luxand provides the face-recognition service. Face Assist is independent of both companies and is not affiliated with either.
This is a prototype intended to demonstrate feasibility. It is not a shipped product, not a medical device, and not (yet) for sale.