MIRRA · How it works

פנים עם רשת המדידה ותקריבי עור — Nine areas, 478 points — and skin close-ups from which texture is measured.

How it works

One selfie, by a window, a few seconds. Beneath it: 478 facial recognition points, nine skin areas, and ten metrics calculated in a scientific color space — each in the areas where it truly reads.

Every line on the face is a measurement, not an estimation. Most of what is said about skin is an opinion that changes with lighting and mood — here, it's a number.

The eye sees "good skin" or "tired skin". The system sees the difference — what moved, in which direction, and from which product, week by week.

No more opinions about your skin — a measurement of it, under the exact same conditions every time, so that two scans taken over time are comparable.

478Facial identification points

9Skin areas

10Metrics

9Unified frames for each metric

How to get a good scan

Everything from the phone already in your hand, by a window, in a few seconds. Three things determine the result — light, distance, and a moment of stability — and all are within your control. This is what a successful scan looks like:

Daylight from a windowFacing a window, even light from both sides. Not with your back to a window, and no warm lamp in a room with daylight.

Hair tied backHair on the forehead or cheeks covers a measured area — and is mistakenly read as a blemish or wrinkle.

No make-upMakeup is measured like skin, and the result would be about the cosmetics, not you. Clean face.

Stand straight, relaxed expressionHead not tilted, no smile — a smile deepens the nasolabial fold and shifts the reading more than anything else in the frame.

Phone at eye levelAt arm's length. Too close or too far, and pores and fine lines disappear in resolution.

Hold still until the ring fillsThe circle around the face fills in a few seconds — it collects 9 frames of the same moment and merges them into one measurement. Movement restarts the fill.

When we decline to measure

This is the part no one puts in an advertisement. An app that always gives an answer seems better, and that's precisely why it's worthless: an answer given from a poor image is a guess with a number next to it.

These are the conditions under which the system says "I will not measure this" and asks again. Each of them is measured, not estimated — and in practice, about 45% of frames are rejected.

No face in the frameWithout a face, there's nothing to measure, and we don't guess from the background.no_face

The image is blurredSharpness below threshold — blur appears to the metric as smooth skin, and this is a false measurement.Laplacian variance < 16.0

Too far awayBelow the required interpupillary distance, pores and fine lines disappear in resolution.Interpupillary distance < 110px

Too dark / too brightPixels stuck at 0 or 255 are not skin — they are erased information.

Head tiltedTilt changes which skin faces the lens, and this changes every color metric.

Eyebrows raisedRaising an eyebrow stretches the forehead and erases real fine lines.> 0.30

A smileA smile deepens the nasolabial fold — the metric would measure the expression, not the skin.

Make-upMakeup is measured like skin. The result would be about the cosmetics, not you.

GlassesThey hide the eye area and what is measured within it.

Hair across the faceHair covers an area, and a covered area is not a measured area.

And why this costs us

Every refusal is an incomplete scan, and a user who needs to try again. A product that sells a good feeling would not do this. A product that sells measurement must — because the only value of a number is that it can be compared to tomorrow's number, and that only works if both were measured under the same conditions.

What exactly is measured, and with what reliability →

← Back to homepage