Face Report: Free Face Analysis, 17 Metrics Measured in 60 Seconds

Axis 1: Symmetry

Symmetry is measured across a vertical midline set by the bridge of the nose and the chin. The measurement uses 11 left-right pairs: eight points down each side of the jaw contour, plus the eye centres, the brow centres and the mouth corners.

For each pair the face report compares how far each side sits from the midline, and turns the total mismatch into a symmetry percentage and a sub-score on the 0–100 scale.

This matters because symmetry is one of the few cross-cultural, cross-method-validated correlates of rated attractiveness in the academic literature. Rhodes (2006) (PMID 16318594) is the canonical review, and its meta-analyses find symmetry attractive in both male and female faces and across cultures.

The face report scores it against the people who scan here. The closer your reading is to 100%, the higher the sub-score, and a typical reading of about 92% lands near the middle of the scale.

In the face report you see the symmetry percentage and its sub-score. If your head is turned or tilted, you see “not measured” and a retake prompt instead of a number, because head angle changes this reading far more than your face does.

Axis 2: Harmony (composite proportion read)

Harmony is how the metrics read together. It does not measure a single distance, and it is not a separate metric or a number in the report. An average of the metric scores would make one outlier and a set of even readings look the same. So the report gives every metric your photo supports (typically 15) its own score.

The report also lists up to three metrics to work on. It ranks the ones you can change yourself (brow grooming, for example) ahead of structural ones like midface ratio, even when a structural reading is lower. This is usually the most actionable finding in the whole face report. The highest-leverage fix is rarely the lowest sub-score, it is the lowest one you can actually change.

Axis 3: Proportion (facial thirds, midface ratio, golden-ratio compliance)

Proportion is three sub-metrics read together in the face report:

• facial thirds (Jefferson 2004; the middle and lower thirds, since the hairline has no landmark to measure)
• midface ratio (eye-line-to-mouth height over brow-to-chin height)
• facial proportion balance, the report's stand-in for golden-ratio compliance (how close three ratios sit to balanced ranges)

All three are computed from the same 68 landmark points and reported as separate sub-scores.

The face report does not score the golden ratio itself. Academic support for the golden ratio as predictive of rated attractiveness is contested, so the proportion-balance metric checks balanced ranges instead of phi (the golden ratio). It is useful as a comparative read across photos of the same person, not as an absolute statement that any specific proportion is required.

For the deeper companion piece, see our golden-ratio facial proportion deep-dive for the literature review, and our evidence-based facial proportion breakdown for the cross-metric proportion picture.

Axis 4: Skin / surface read

The face report has no skin score, and none of its 17 metrics reads skin. It does not run a clinical skin-condition diagnosis either (that requires a trained clinician, controlled lighting and dermoscopy, which is magnified skin imaging).

What it reads from the photo surface is whether the photo can be measured. It checks overall brightness, exposure on the face itself, and whether one side of the face is lit much more strongly than the other.

A photo that fails those checks is flagged before the scan, because it makes the landmark readings less reliable, not because it lowers the number. This is a deliberate boundary: a single photo is not enough information to judge skin condition, and we will not pretend otherwise.

Axis 5: Photo conditions (what the face report needs to measure you)

There is no smile score in the face report. None of its 17 metrics reads the smile or the teeth, and a smile works against the report: on a smiling or parted mouth, lip ratio is marked not measured. What the report needs is a photo it can measure:

• Face the camera straight on with your head level (a turned or tilted head means symmetry is not measured).
• Keep your mouth relaxed and closed (a smile means lip ratio is not measured).
• Keep your eyes relaxed (the hunter eye index changes with how open they are).
• Hold the camera at eye level (midface ratio shifts with camera pitch).

These conditions decide whether a metric can be measured and whether two scans are comparable. They are not a proven way to raise the number: in our data, better-lit, sharper photos did not measurably move the same person's score. The step our data does support is taking two or three photos and reading the middle one, because a single photo is a thin sample of your face.

Symmetry research and the plateau effect

Facial symmetry is one of the more replicated correlates of rated attractiveness in the academic literature. Rhodes (2006) (PMID 16318594) is the canonical review. Its meta-analyses find averageness, symmetry and sexual dimorphism (how typically male or female a face's features are) all attractive, in both male and female faces and across cultures.

The face report does not assume a plateau effect, where extra symmetry stops counting past some point: scored against the people who scan here, the sub-score keeps rising all the way to 100% symmetry.

Measured symmetry is also partly a property of the photo: turning or tilting the head shifts the landmarks more than real asymmetry does. That is why the face report does not measure symmetry at all on a turned or tilted head, and asks for a straight-on retake instead of printing a number it cannot stand behind.

Averageness, FWHR, and dominance signaling

Averageness (a face whose proportions cluster near the population mean) is associated with higher rated attractiveness in multiple cross-cultural samples (Langlois & Roggman 1990; Rhodes 2006).

FWHR (facial width-to-height ratio) is a separate marker, tied in the research to aggressive behaviour in men (Carré & McCormick 2008, PMID 18713717) and to judgments of dominance and threat (Geniole et al. 2015, PMID 26181579). The face report measures both and shows each as its own metric, so you can read the averageness inputs (facial thirds, midface ratio) and FWHR separately.

This separation matters because averageness and dominance are different reads. A face can be high-FWHR and average-proportion at once, or high-averageness with low-FWHR. The face report surfaces both reads separately so the user can decide which one is relevant to the photo context. A LinkedIn headshot, a dating-app lead photo and a casting profile picture each weight these reads differently.

Marquardt mask and the golden ratio (a hedged inclusion)

Marquardt (2002) proposed a phi-based facial mask aligning twelve landmark points to ratios derived from 1.618. Marquardt-mask alignment is widely cited in cosmetic and craniofacial reference work, but empirical support for the golden ratio as a strong predictor of rated attractiveness is contested.

Holland (2008, PMID 18175168) found the mask best describes the facial proportions of masculinized white women as seen in fashion models, is ill-suited to non-European faces, and does not match most people's preferences.

The face report does not score golden-ratio compliance. Its nearest metric, facial proportion balance, checks three ratios (height-to-width, midface-to-lower-face, eye-span-to-face-width) against balanced ranges instead of phi. It is informative as a comparative read across photos of the same person, not as a settled predictor.

We document this so the user knows where the evidence is strong and where it is partial.

First impressions, expression, and what the report leaves out

First impressions of a face form fast. Willis & Todorov (2006, PMID 16866745) found that judgments of attractiveness, likeability, trustworthiness, competence and aggressiveness made after a 100-millisecond look at a face correlated highly with judgments made with no time limit. That is the reason the photo matters, and the reason this page treats the score as a read of one photo rather than of you.

The face report does not score expression. None of its 17 metrics reads the smile, and a smile works against the measurement: on a smiling or parted mouth, lip ratio is marked not measured. The geometry is most readable on a relaxed, closed mouth, which is also what keeps two scans of the same person comparable.

A score above 80 is associated with strong photo signals, not a guarantee

A reading this high is the least settled kind in our data. We looked at the 281 people whose Face Score landed in the top quarter (75 or higher) and who got a different number on a second photo the same day. Of those, 74% came back down, and the median change was 8 points down.

It does not guarantee any specific outcome. Match rate, callback rate and partner-selection results depend on context (the platform, the audience, the bio, the other photos in the set) that the face report cannot see from a single image. Treat a score above 80 as a green light to use the photo, not a promise of any particular result.

A score in the 60 to 80 band usually flags two or three fixable axes

The face report lists up to three metrics to work on, the ones you can change yourself ahead of the structural ones, rather than telling you to fix every weak metric.

A reshoot is not a shortcut to 80. In our data, 796 people had a first Face Score of 60 to 79 and took a second photo the same day. Of those, 99 reached 80 on it, and more went down than up.

A score in the 40 to 60 band is a read of this photo, not a verdict on your face

The hedged read is that the face report cannot tell you a 40 to 60 score reflects your face, only that it reflects this specific photo of your face. In our data, a second photo the same day was a coin flip from here. Of 437 people whose first Face Score was 40 to 59, 144 went up, 145 went down and 148 got the same number.

Better-lit, sharper photos did not measurably move the same person's score in our data either. The useful step is not a better-lit retake but more samples: scan two or three photos and read the middle one.

A score below 40 almost always means re-shoot before drawing any conclusion

A single photo is a thin sample of a face. In our data, a person's Face Score moved a median of 4 points between their first photo of the day and the next one. About 1 in 9 of those pairs (11.2% of 2,223) moved 30 points or more. Across all of one person's photos from the same day, the gap between the highest and lowest Face Score is wider: a median of 11 points.

A number this far from the middle is weak evidence on its own. The face report recommendation is to re-shoot first, straight on with your mouth relaxed and closed, and read the middle of two or three scans before interpreting any one of them.

What the score is not predictive of

The face report is a single-photo, 17-metric photo read. It is not predictive of long-term partner selection, relationship satisfaction, professional outcomes, or anything that depends on context the face report cannot see.

Research on the first-impression literature is clear that initial reads explain a portion of variance in rapid-judgment tasks. The same research shows they explain a much smaller portion of variance in actual relational or professional outcomes (which are dominated by behavior, not appearance). The face report is a useful comparative tool for picking a photo. It is not a useful tool for predicting your life.