Free AI attractiveness test

The test maps facial landmarks and calculates bilateral symmetry, canthal tilt, facial width-to-height ratio, an approximate visible gonial angle, interocular ratio, and the balance between the middle and lower facial thirds. Those values are combined into a PSL-style score.

The geometry is real; the conversion into one attractiveness number is a product heuristic. No accepted scientific formula turns a face into a universal beauty score, and a still photograph cannot capture expression, voice, familiarity, styling or cultural preference.

1. 1

   Choose a clear photo

   Use one front-facing face at eye level with even light and a neutral expression.

2. 2

   Run the model locally

   The face-landmark model reads the image in browser memory; the image and landmarks are not sent to a server.

3. 3

   Inspect the score

   See the score and three free measurements, then decide whether the full six-metric report is useful to you.

Accuracy has two separate meanings here. Landmark placement can be consistent on a clear photograph, while the attractiveness interpretation remains subjective. A tilted head, close wide-angle selfie, smile, crop or hard shadow can alter visible geometry even when the person has not changed.

• Compare only photos taken under similar conditions.
• Use the individual measurements as descriptive signals, not diagnoses.
• Do not treat the number as a population rank or statement of personal worth.
• Close the page if appearance scoring is making you feel worse rather than curious.

Start with the measurement rows rather than the headline number. Each row names a geometric feature, shows the value detected in the photograph and explains the reference used by this calculator. A larger overall score only means that the selected measurements sit closer to this product’s chosen targets. It is not evidence that every viewer would prefer the face, and it does not turn taste into a laboratory measurement.

Look for patterns instead of fixating on the smallest decimal difference. A low-confidence photograph can move several measurements together: yaw affects visible face width and symmetry, pitch changes vertical proportions, and a smile shifts the lower face. When one result surprises you, repeat the photograph under controlled conditions before drawing a conclusion. The repeat is a quality check on the image, not a challenge to obtain a preferred score.

• Read the disclosed metric definition.
• Check whether the photo met the capture guidance.
• Compare only results made by the same calculator.
• Treat small differences as photo-sensitive.
• Keep the number separate from self-worth.

A phone camera can be technically sharp and still be poor measurement input. A close selfie exaggerates the nose and center of the face while compressing the sides. Portrait mode may alter edges, beauty filters can move landmarks, and overhead light can hide one jaw corner. Step back, use modest zoom if the camera allows it, keep the lens level with the eyes, and leave the entire face inside the frame.

The detector also needs one visible face. Group photographs, sunglasses covering the eye corners, hair across the jaw, extreme expressions and partial profiles can prevent a stable result. Rejection is preferable to a confident-looking score from unusable input. Omogle explains common capture problems so you can correct the photograph without sending it anywhere or asking a stranger to inspect it.

• Use one adult face in the frame.
• Remove filters and face-warping effects.
• Keep both eye corners and the jaw visible.
• Choose soft, even front light.
• Avoid screenshots of compressed social-media images.

The calculator is designed for private curiosity about one image, not public competition. It has no leaderboard, social profile, stranger matching or database of submitted faces. That product boundary matters: a measurement tool can explain its own formula, while a social ranking system encourages comparisons that the underlying geometry cannot justify.

Facial attraction also changes with context. Expression, movement, familiarity, voice, personal style and cultural expectations are absent from a still image. Even a perfectly implemented landmark model cannot recover information that was never present in the photograph. The honest output is therefore narrow: a repeatable description of selected proportions, accompanied by explicit limits.