Facial harmony is the score for whether the parts of a face agree with each other. Not whether the nose is good, but whether that nose belongs on that face, next to those eyes, above that jaw.
Across 17,032 AI face analyses it averages 5.58, almost exactly the overall average of 5.59. That coincidence is not a coincidence, and unpacking it explains both what harmony measures and why it behaves so strangely compared to every other feature on the report.
Two numbers that look contradictory
The first: harmony correlates 0.99 with the overall face rating. That is the highest figure of the twelve measured features, higher than bone structure (0.96), higher than symmetry (0.95), higher than the jawline (0.92).
The second: harmony is almost never anybody's best feature, at 0.3%, and almost never anybody's worst, at 0.5%. It is the lowest figure in the dataset on both counts simultaneously. Nothing else manages that.
| Feature | Correlation with overall | Best feature | Worst feature |
|---|---|---|---|
| Facial harmony | 0.99 | 0.3% | 0.5% |
| Nose | 0.97 | 0.5% | 9.0% |
| Bone structure | 0.96 | 3.2% | 2.6% |
| Chin | 0.96 | 0.1% | 8.4% |
| Cheeks | 0.96 | 1.0% | 6.6% |
| Symmetry | 0.95 | 3.9% | 2.6% |
| Eyes | 0.95 | 20.1% | 0.1% |
| Forehead | 0.93 | 2.4% | 15.0% |
| Jawline | 0.92 | 16.0% | 9.1% |
| Lips | 0.90 | 11.4% | 6.2% |
| Eyebrows | 0.90 | 29.3% | 1.9% |
| Skin | 0.87 | 11.8% | 38.0% |
Read down the table and every other feature has a personality. Skin is the great differentiator: lowest correlation with the overall score, but a weak point for 38.0% of people. Eyebrows lead the best-feature column at 29.3%. The chin almost never wins and often loses. These features stick out in one direction or the other.
Harmony never sticks out. It tracks the overall score more tightly than anything else while being, for practically everybody, neither their strongest nor their weakest number.
Why those two facts are the same fact
A feature can only be your best or your worst if it sometimes departs from the rest of your face. Harmony does not depart, because harmony is a reading of the rest of your face.
The other eleven scores measure a region: this jaw, this nose, this skin. Harmony measures the relationship between them. If your features are all around 5.5, your harmony is around 5.5, because a set of mutually consistent middling features is exactly what a middling harmony score describes. It cannot be your standout, because there is nothing for it to stand out from.
Which means harmony is not really a thirteenth measurement. It is closer to the overall score restated in the language of proportion. A 0.99 correlation is what a restatement looks like in data.
This is also why harmony is marked barely improvable. Nothing acts on harmony directly. It moves only when the things underneath it move, and then it moves nearly in lockstep with the overall rating.
How harmony differs from symmetry
This distinction gets blurred constantly, and the two scores answer genuinely different questions.
Symmetry is a physical property. It asks whether the left half of the face matches the right when one is mirrored across the midline, and on FaceRating.ai the model estimates it by looking at the photo rather than by placing landmarks. It exists independently of every other feature: you can assess it without knowing whether the nose is well proportioned or whether the jaw suits the cheekbones. Its 0.95 correlation with the overall rating is the strongest of any real, independent, physically measurable feature - a genuine predictor, established as such well beyond our dataset. The face symmetry test scores it on its own and explains what a symmetry score does and does not tell you.
Harmony is a judgement about a relationship rather than a single physical property. It has no midline to compare across and no one distance to estimate. It asks whether the parts cohere, which is a question about the whole set at once. Its 0.99 correlation is higher than symmetry's, but it is higher for a different reason: symmetry predicts the score, harmony largely summarises it.
Both statements hold at once, and it is worth being precise about them. Symmetry is the strongest independent predictor. Harmony is the tightest summary. A predictor and a summary are not competing for the same title.
The distributions back the distinction up. Symmetry has a standard deviation of 1.01 and a high-to-low gap of 3.81 points; harmony sits at 1.07 and 4.11. Both spread people out more than the tightly-packed features like the forehead (0.78) or the nose (0.93). Harmony has range. What it does not have is independence.
Why fixing one feature so often changes nothing
This is the practical consequence, and it is the reason the harmony score is worth having on a report at all.
If harmony is about agreement between parts, then improving one part in isolation can leave the agreement unchanged, or occasionally reduce it. A feature that is now noticeably stronger than everything around it is, by definition, less consistent with the rest of the face.
The dataset shows how narrow the margins are. The average distance between a person's best and worst feature is 1.38 points. 44.8% of people tie for their best feature and 58.0% tie for their worst, meaning most faces do not have one clear weak link at all - they have a band, and the band moves together or not at all.
Compare that to what a high overall score looks like. Among the 560 people scoring 8.0 or above, the twelve feature averages run from 8.90 down to 7.66. The spread is 1.24 points, roughly the same narrow band, just relocated upward. High scores are not one spectacular feature carrying ten ordinary ones. They are twelve features sitting high together, which is another way of saying they are harmonious.
The honest implication is unwelcome but useful: the routes that move a whole band are the ones that matter, and they are few. Skin and grooming affect how the whole face reads. So does body composition, hair, and photograph quality. Chasing a single feature usually does not move the band, and the harmony score is the number that tells you so.
Reading your own harmony score
Treat harmony as a cross-check, not as a thirteenth thing to fix. If your harmony score is close to your overall score, which for most people it will be, it is confirming that your features are consistent with each other. If it sits noticeably below your feature averages, that points at proportion rather than at any one region - and proportion is where hairstyle, framing and photo angle do their work.
Only 12.5% of people score 7 or above on harmony, and the average is 5.58, so a score in the mid-fives here is the ordinary result rather than a sign of anything wrong.
To see the number itself, the facial harmony test scores proportion and balance on its own. For the full twelve-feature breakdown that harmony summarises, run the face rating tool, or take the attractiveness test to get the same score placed against the distribution quoted throughout this article.