FIELD NOTES / FACE MEASUREMENTS
Understand a face rating
Read the reference rules before deciding what a number means.
AI-assisted product documentation, checked against our published method and the technical sources below. Diagrams and worked numbers are illustrative, not personal test results or a clinical review.
Separate the photo from the person
The displayed number describes how a set of visible two-dimensional measurements compares with this product’s reference rules. It does not summarize how every person will perceive you. Expression, framing and detected point placement all affect the input. A different image can produce a different result, even when no physical feature has changed. The method page and result panel make that context available alongside the score.
Read the four components
The overall score combines symmetry, visible-landmark thirds, facial fifths and a visible face proportion with fixed weights. The ten-point rating is the same overall result divided by ten, not a second independent opinion. Read each component’s raw measurement and formula. A lower component identifies a reference mismatch in this photograph; it does not establish a flaw or tell you that a change is necessary.
Use the limits as part of the result
A provisional result means the photo triggered a caution. A rejected result has no overall score. Missing dimensions are not quietly replaced or assigned perfect values. Compare results only when they use the same method version and suitable photo conditions. If a number makes you feel worse without helping you understand anything useful, leave the number aside and focus on practical choices you already enjoy.
Read a result in this order
A useful result explains its inputs before suggesting an interpretation. Start with the image quality and the measurement definitions. The headline number is only one summary of the selected rules, so it should be the last thing you compare.
- Check whether the result is scored, provisional or unavailable. A provisional result carries a photo-quality caution; an unavailable result has no total.
- Read the dimension labels. Visible thirds start at a detected mesh point, not a known hairline; horizontal fifths are a geometric reference.
- Open the measurement details to see raw ratios, point definitions, formulas and method version. Similar labels on another website may use different endpoints.
- Look for the reason behind a lower dimension. A reference mismatch describes this calculation; it is not proof of a flaw.
- Use a consistent second photograph if you want to compare captures, then inspect dimension changes before the total or ten-point rating.
Worked example: how the total is formed
Suppose four valid dimension scores are 90, 80, 70 and 60. These are invented teaching numbers, not readings from a person. The table shows each contribution using the current product weights. This calculation demonstrates arithmetic only; it does not validate the choice of weights.
| Dimension | Illustrative score × weight | Contribution |
|---|---|---|
| Mirror geometry | 90 × 0.30 | 27.0 |
| Visible thirds | 80 × 0.25 | 20.0 |
| Horizontal fifths | 70 × 0.25 | 17.5 |
| Selected proportion | 60 × 0.20 | 12.0 |
The illustrative total is 76.5/100, or 7.65/10. If the proportion score alone increased by ten points while the other inputs stayed fixed, the total would increase by two points. This is a property of the formula, not a prediction that a styling change will improve a face. Missing dimensions are not reweighted into a new total.
Compare a small series honestly
- Repeat one exact file three times using the same method version. Record the quality status and each returned dimension, including any missing result.
- Next compare three separately captured neutral photos with similar conditions. Keep file labels so you can tell an identical-file repeat from a new photograph.
- For each dimension, compute range = maximum minus minimum among valid readings. A small range means those particular runs were close; it does not establish accuracy against a true value.
The bands describe a weighted reference match. They are not population percentiles: 76.5/100 does not mean more attractive than 76.5% of people. If a score provides no useful information, you can stop and use the photo or style guides without treating the number as a goal.
Use this with the right tool
Sources and scope
These sources explain detection and photography. They do not validate our weights, reference targets, attractiveness labels or styling suggestions. Our own scoring rules are documented on the method page.
- Google AI Edge · Face Landmarker overview
Documents the landmark model and its outputs; it is not a beauty-rating model.