When a face analyzer draws dots over a portrait, it can look as though the dots are the analysis. They are only its starting coordinates. A landmark model estimates where features appear in the image; a separate set of calculations decides which distances, alignments, and ratios to report. Keeping those steps apart makes an AI face analysis easier to understand.
What is a facial landmark?
A facial landmark is a coordinate associated with a feature or part of the face surface in an image. Depending on the model, points may follow the eyes, brows, lips, nose, cheeks, jaw, and forehead. A dense collection of points is often called a face mesh. The dots are estimates produced from pixels, not physical markers attached to your face.
Landmarks are useful because they let software compare positions in a consistent coordinate system. For example, it can ask whether corresponding eye points sit at similar heights, or what share of a visible face height lies between the brow and nose base. A point by itself is not a beauty score.
Where do the points appear?
The simplified illustration highlights regions that are relevant to the measurements discussed below. Real landmark models return many more points, and the exact indices and definitions depend on the model.
ILLUSTRATIVE LANDMARKSHow does a photo become an analysis?
- Find the face. The model locates a face in the image. A photo with no face or more than one face is unsuitable for our single-person test.
- Estimate landmarks. The model returns point coordinates across the detected face. These are positions, not interpretations of personality or attractiveness.
- Check the input. Our tool checks angle, face size, brightness, and sharpness before scoring. No face, multiple faces, excessive side angle or a face too small in the frame stops the test. Brightness and sharpness warnings can instead ask you to retake the photo or confirm that you want to continue.
- Calculate relationships. Selected points become distances, proportions, and left-right comparisons. The product then applies its published, versioned scoring weights.
What can landmarks tell our analyzer?
Our free geometric score uses selected landmarks for four visible relationships. The full report adds other descriptive readings and styling suggestions. Each is a description of this photo under the current method, not a universal ranking of the person in it.
- Symmetry: corresponding landmark pairs are compared around a face midline after the image coordinates are leveled to the eye line.
- Visible facial thirds: vertical distances from a visible forehead point through the brows and nose base to the chin are expressed as three shares. The upper point is not the hairline. See the full facial-thirds explanation.
- Facial fifths: selected horizontal points across the eye and face-width region define five width segments for comparison.
- Face length-to-width balance: the visible top-to-chin distance is compared with the measured face width.
The report shows component scores, their weights, and underlying readings. Those details matter more than treating one total number as a complete account of a face.
Why can landmarks move between photos?
The same person can produce different landmark coordinates in different pictures. Turning or tilting the head changes the projected shape. A close camera changes perspective; strong expression moves the mouth and brows; blur, shadows, filters, or hair can hide useful boundaries. A model may still return points on a difficult image, but returned points are not proof that the measurement is dependable.
Use a clear, front-facing image at roughly eye level, with even light and a relaxed expression. If you want to compare results, keep the camera setup similar. Our photo guide explains how to prepare that input.
What does the confidence percentage mean?
The percentage on our result page combines frontal-position quality, sharpness and relative face size. It is an app-defined suitability indicator, not a probability that every point is correct or that the final score predicts attractiveness. Lighting warnings are checked separately. A high percentage is still a reason to inspect the overlay, especially near hair, glasses or strong shadows.
Point detection, quality review and scoring are separate steps. The methodology page documents each one, including the score version and component tolerances.
What can facial landmarks not tell you?
Points on a single image cannot establish someone's identity, health, personality, intentions, or worth. Nor do they capture voice, movement, style, social context, or the many subjective ways people perceive attractiveness. Our tool does not use its landmark output for identity recognition, emotional diagnosis, or medical advice.
A geometric reference is also a choice made by a product. Different landmark definitions, photo conditions, target values, and weights can produce different scores. That is why we name the scoring version and avoid claiming that a score is a population percentile. For a broader discussion, read How Accurate Can an Online Attractiveness Test Be?
Privacy and further reading
The free model runs in your browser. You can share a derived card, while the paid full report generates a PDF after payment confirmation and emails it to you. Opening a full-report preview saves a temporary local browser result rather than an account history. Read the privacy policy for the processing and retention details.
For the underlying landmark technology, consult Google's MediaPipe Face Landmarker documentation. Google's Face Detection Concepts offers additional background on detected facial features. Those resources describe model capabilities; the four measurements and scoring choices above describe our product.
Explore a photo-based analysis.
Select a clear front-facing photo to see the measurements behind your geometric result.