What one photograph can measure — 51 values, and 13 it cannot reach

Almost every face analysis tool lists what it measures. Almost none lists what it cannot — which is the part you need before you let a number influence a decision.

Last updated 2026-09-21

Start here: a photograph carries no scale

Nothing inside a single photograph says how large anything in it really is. A camera held closer makes a face fill more of the frame without the face changing at all. Without an object of known size sitting in the same plane as the face, there is no way to turn pixels into millimetres.

The consequence is direct: values measured from a photograph are ratios and nothing else — "nose width relative to face width", never "a nose 38 millimetres across". A tool that reports millimetres from one photograph is guessing your camera distance without telling you.

A ratio is not the poor relation of a real unit. Facial proportion is a question about how parts relate to each other in the first place. But knowing a number is a ratio tells you what it can and cannot be compared against.

51 values measured, out of a catalogue of 87

DooDee measures 51 facial ratios and angles from anatomical landmarks detected in the image. The catalogue of characteristics the product has an opinion about holds 87 entries. Of those, 74 have something real behind them and 13 are marked not measured, each with its reason.

Those 13 are not a backlog. They are the entries a single two-dimensional photograph cannot answer in principle, not ones awaiting code. Showing them rather than dropping them is what stops the list looking more complete than it is.

12 can be compared with other people. The rest compare you with yourself

This is the part usually skipped. Saying a value is "above average" requires a published mean and standard deviation for a comparable population, and in the literature that actually exists, there are not many.

Of the 51 measured values, 12 have published reference figures to be scored against, drawn from a cohort of 240 Thai adults aged 18–35. Everything else compares your face with itself — left against right, one region against another, or this scan against your last one.

The distinction matters when you read a result. "Above the reference range" and "larger than the other side of your own face" are entirely different sentences, and only the first one is about anyone else.

The same applies to percentiles. A figure computed from a handful of people is not a percentile, so one is not presented as fact until at least 30 scores exist behind it.

Some values differ from the textbook systematically — and that is not an error

Landmarks detected in an image and the points anatomical texts use are not always the same points. When the points differ, the number differs in a fixed direction rather than randomly. The two clearest cases:

Value	Why it differs
Cheek width ratio	The denominator is the distance between the detected cheek landmarks, which sit wider apart than the zygion points used anatomically. The result runs systematically below the 0.45 figure usually quoted.
Nasolabial angle	Taken from the landmark at the crease below the lip, which is shallower than the clinical point B. The result runs systematically above the 120–130° range cited in textbooks.

Both values can track change in the same face over time, because the method is identical every time. Neither may be compared directly against the textbook figure. A tool that sets a number like this beside 0.45 and tells you that you fall "below standard" has made the wrong comparison before it started.

What one front-facing photograph cannot tell you

Where the eyelid crease sits — the detected upper-eyelid points follow the lash line, not the crease, so there is no crease data to measure.
The true jaw angle — that is read from a lateral radiograph. A frontal photograph can only say how square the jaw looks from the front.
How deep an under-eye hollow is — what is measurable is shadow intensity relative to your own cheek, and a light source above you produces the same reading as a genuine hollow.
The cheekbone apex in profile — it does not sit on the midline, so in a side view it falls on the far side that is not actually visible. The value is an estimate.
Brow shape — the detected brow points follow the orbital bone ridge, not the shaved or drawn edge of the eyebrow.
Jawline definition — this is the sharpness of a shadow edge, so it moves with the direction of the light more than with the shape of the jaw.

One more worth stating plainly: the symmetry figure measures left-right difference across a set of landmarks. It is not a symmetry score for the whole face, and a few degrees of head rotation will move it. If your symmetry value shifts between two scans, the most likely explanation is the camera angle, not your face.

"Harmony" is not a beauty score

The value named harmony combines several proportion measurements under a fixed formula. It describes how the proportions of your face relate to one another. It does not describe how attractive your face is. Attractiveness is a matter of taste and culture, which no formula stands in for, and a tool handing out a "beauty score out of 100" is presenting an opinion in the shape of a number.

In practice these figures are most useful against the question "what about my face is different from what I assumed", and least useful against "what did I score".

So how should these numbers be used

As the start of a question rather than the end of one. Measurements from a photograph let you walk into a consultation knowing what to ask about, instead of starting from nothing — or from whatever an advertisement told you to fix.

And as a comparison with yourself over time. A value captured the same way each time describes your own change far more reliably than it describes where you stand relative to anyone else.

Analyse your own face