Your face stays yours
There is no image server to upload to. Frames are analysed in memory on your device and discarded. Not a policy — an architecture.
An independent app, built in the open.
Glassly is made by one person — not a lab, not a conglomerate. That shapes everything about it: what it measures, what it refuses to collect, and who it answers to.
Sahaj Sinha
Founder · Designer · Engineer
- Based in South Korea
- Every line of code, first to last
- Self-funded, no investors
Why I built Glassly
Skincare advice is loud, contradictory and usually trying to sell you something. I wanted the opposite: a way to look at your own skin, get a consistent read on it, and watch whether what you're doing is actually working — without handing your face to a server.
So Glassly measures rather than guesses. Five metrics across five regions of your face, the same way every time, so week-to-week change means something. Then it explains what it found in plain language and builds a routine around it.
The hardest constraint was self-imposed: no photo may ever leave your phone. That ruled out every convenient cloud API and meant shipping the machine-learning models inside the app itself. It made the build far harder. It's also the part I'm least willing to compromise on — your face is not training data, and it isn't a product.
— Sahaj Sinha
Four rules I don't break.
Every product decision gets checked against these.
Measure, don't flatter
The score reflects what the models actually read. A number that only ever goes up would be comforting and completely useless.
Guidance, not diagnosis
Glassly is a cosmetic tool. It will never claim to detect a medical condition, and it says so plainly rather than hiding behind a disclaimer.
No dark patterns
Three real scans free, no card. Cancel in two taps from your account. No countdown timers, no fake scarcity, no ads, no data sold.
How it actually works.
No magic, and no cloud. Here's the honest version.
01
Detect & align
On-device vision models find your face and map it precisely, so the same regions are compared every scan. If lighting or stability is poor, Glassly waits rather than recording a bad reading.
02
Segment the skin
A segmentation model separates skin from hair, background and clothing, so a dark jumper or a shadow can't quietly drag your numbers around.
03
Measure five metrics
Hydration, texture, redness, oiliness and clarity are computed per region from the pixels themselves, then combined into one score. All of it runs on your phone's own processor.
04
Explain & recommend
Readings map to concerns, concerns map to ingredients, and ingredients map to an AM/PM routine — with conflicts flagged, so you're never told to layer two things that fight each other.
Accounts and subscription state use Firebase; payments are handled by Polar as our merchant of record. Open-source components are credited in the app under Settings → Open source licenses.