URATEN

Facial measurement and scoring methodology

URATEN provides a detailed, measurement-led analysis of facial proportions visible in suitable frontal and side-profile photographs. It turns reviewed facial landmarks into named ratios, angles, reference-line relationships, symmetry measures, individual metric scores, and a model-based overall summary.

We truly care about giving each user the most accurate and useful result that their photos can support. That is why the system checks photo quality before analysis, lets the user review important landmarks, keeps measurements separate from their scores, and clearly explains where photographic analysis has limits.

URATEN does not ask a large language model or multimodal model to look at a photo and subjectively choose an attractiveness score. The scoring pipeline is based on defined geometry and deterministic calculations. The same approved inputs and scoring version produce the same output.

1. What URATEN measures

URATEN is a two-dimensional facial geometry analysis system. It measures relationships between defined points on the visible face, such as jaw width relative to cheekbone width, the balance of the vertical facial thirds, eye and mouth spacing, profile angles, and the position of features relative to constructed reference lines.

This is fundamentally different from asking an AI system for an impression of the whole face. Every reported measurement comes from coordinate relationships that can be inspected and, where the interface allows, corrected before analysis. The result is a structured account of specific proportions rather than an unexplained visual opinion.

The core analysis follows a clear sequence:

  1. The browser checks whether each photo is suitable for the requested view.
  2. An on-device model detects frontal facial landmarks, and the user reviews key points.
  3. The user places the defined points required for an optional side profile.
  4. URATEN normalizes the approved coordinates and calculates named geometric measurements.
  5. Each measurement is evaluated with its configured reference function.
  6. Fixed weights combine the available metric scores into view-specific results and an overall summary.

With suitable frontal and profile inputs, the current report can include up to 40 named metrics. The exact measurement remains distinct from the score assigned to it. The measurement describes the geometry derived from the approved landmarks. The score describes how the current URATEN model interprets that value relative to its configured reference.

This separation is important. It means users can examine the underlying proportion even if scoring references evolve, and it makes the basis of the analysis much easier to understand than a single score with no supporting breakdown.

2. Photo quality comes before analysis

A reliable measurement begins with a suitable photograph. Perspective can change apparent widths, angles, projection, and symmetry, so URATEN checks pose in the browser before any landmarks are submitted for scoring. We would rather ask for a better photo than produce a precise-looking result from an unsuitable one.

The current checks include the following safeguards:

  • Face detection: the requested view cannot proceed unless the browser detects a face.
  • Frontal pose: a detected yaw greater than 15 degrees is rejected, and the user is asked to provide a straighter photograph.
  • Side profile pose: an optional profile photo must show at least 60 degrees of yaw, so a three-quarter view is not treated as a true profile.
  • Symmetry eligibility: symmetry is reported only when detected frontal yaw is below 5 degrees. When pose is outside that tighter range, symmetry is marked unavailable and its weight is removed from the frontal calculation.

These checks protect the integrity of the analysis, but pose is not the only part of a good capture. For more repeatable measurements, users should choose even lighting, a neutral expression, an unobstructed face, a level camera, and enough camera distance to reduce close-range lens distortion. Comparisons over time are most informative when photographs are taken under similar conditions.

Automated quality checks are designed to catch important problems, not to claim that every accepted image is perfectly calibrated. Blur, expression, lighting, occlusion, lens choice, and landmark visibility can still affect the result. The landmark review step gives the user another opportunity to replace a poor capture or correct a misplaced point before continuing.

3. Privacy-first photo processing

We care deeply about the privacy of the people who trust URATEN with an analysis. Photo preparation, face detection, and initial landmark detection run locally in the user's browser. Raw photo pixels are not uploaded to URATEN's servers.

After the user gives consent and reviews the landmarks, the browser sends the numerical coordinates and required analysis settings to the analysis endpoint. The server uses those coordinates in memory to calculate the requested metrics and scores, then discards the raw coordinates. They are not written to the database.

Account features may store the resulting facial measurements and analysis so the user can access them again. URATEN does not use landmark coordinates or derived facial measurements to identify a person, authenticate an identity, search a face database, or match one person's face with another. The Privacy Policy explains the data lifecycle, consent process, retention, and user rights in detail.

4. Landmark detection and review

For a frontal view, URATEN runs MediaPipe Face Landmarker on the user's device. This model detects a dense facial mesh. URATEN then uses a defined subset of those points for features such as the eyes, pupils, brows, nose, mouth, cheekbones, jaw, and chin.

The use of machine learning at this stage is limited to locating facial landmarks. It is not an attractiveness judgment. The detector does not choose a score, decide which features are appealing, or change the scoring rules.

Relevant frontal points are displayed over the photograph before scoring. The user can drag key points to improve their placement and must verify the hairline point, which cannot be inferred reliably from the facial mesh alone. For the optional profile view, the interface guides the user through placing 13 defined anatomical reference points after the pose check.

This review step matters because even a small coordinate error can affect a narrow angle or ratio. Showing the landmarks makes the process inspectable and gives the user meaningful control over the input. It also keeps the scope honest: reviewed points from a photograph can support useful geometric analysis, but they are not a substitute for calibrated three-dimensional or clinical imaging.

5. From landmarks to facial measurements

Before measuring, URATEN normalizes orientation where appropriate. Frontal coordinates are rotated so the inner eye corners are level. Profile coordinates are converted to a consistent facing direction and aligned to the Frankfort reference plane using the porion and orbitale points.

The engine then calculates relative distances, ratios, angles, signed distances to reference lines, and symmetry deviations. Most results are ratios or angles rather than physical lengths. This makes the analysis less dependent on image size while avoiding a claim that an ordinary, uncalibrated photograph can provide exact millimetre measurements.

Measurement groups

  • Frontal proportions: face width-to-height ratio, bigonial-to-bizygomatic ratio, facial thirds, mouth-to-nose ratio, lip ratio, and cheekbone height.
  • Eyes and feature spacing: canthal tilt, eye aspect and spacing ratios, brow-eye distance, eyebrow tilt, interpupillary relationships, and nose-tip deviation.
  • Jaw and profile angles: gonial and mandibular plane angles, chin projection, facial convexity, nasofrontal and nasolabial angles, and ramus-to-mandible ratio.
  • Profile lines and projection: E-line, S-line, Burstone line, orbital vector, Z angle, nasomental angle, nasal projection, facial depth, and nasal-tip rotation.

These measurements give the analysis its practical depth. A general impression such as “balanced” is difficult to inspect. A named value identifies what was measured, which points produced it, and how it relates to the current reference model. Users can therefore understand why a metric contributed positively or negatively instead of being asked to trust a hidden judgment.

6. How metric and overall scores are calculated

Individual metric scores

Each geometric value is compared with either a target point or an accepted interval in the current scoring model. Some reference parameters vary with the selected male or female analysis setting, while others are shared.

A Gaussian function is used when the model has a single target. The metric receives its highest score at that target and decreases smoothly as the measured value moves farther away:

s = 10 × exp[-0.5 × ((x - μ) / σ)²]

In this function, x is the measured value, μ is the configured reference, and σ controls how quickly the score changes with distance from that reference. A split-Gaussian can apply different tolerances below and above a target. A plateau function gives the maximum metric score throughout an accepted interval, then decreases smoothly outside it.

Smooth functions avoid arbitrary pass-or-fail edges and show degrees of difference. Given the same approved landmarks, selected analysis setting, available views, pose estimate, and scoring version, the engine produces the same result. This repeatability is one of URATEN's core strengths: the rules are applied consistently, and no model improvises a rating from the photograph.

The overall score

Metric scores are combined using fixed weights. In an analysis with both suitable views, the current model gives 55 percent of the combined score to frontal metrics and 45 percent to profile metrics. Within each view, higher-weight metrics contribute more. If only a frontal view is available, the frontal weights are normalized to produce the summary on their own.

The overall attractiveness score is a compact summary of how the selected measurements fit URATEN's current reference model. It is not, and is not presented as, an objectively or perfectly accurate verdict on attractiveness. Human perception also reflects personal preference, culture, expression, movement, styling, and many qualities that facial geometry cannot fully represent.

We believe the most responsible and useful way to present the score is as a navigation aid within a much richer report. The exact measurements, metric-by-metric scores, view-specific breakdowns, landmarks, and contribution patterns explain the result. Two people can reach a similar overall score through very different combinations of measurements, and URATEN preserves that detail instead of hiding it behind one number.

Percentile-style rating labels are mathematical transformations of the internal metric score under the model's distribution assumptions. They are model-based comparisons. They do not claim that URATEN directly surveyed or ranked the user against a live, representative population.

7. What the analysis is useful for

URATEN's clearest value is a structured, inspectable description of facial proportions visible in a controlled photograph. The detailed analysis remains useful even when a user chooses to treat the overall score as secondary.

  • A numerical baseline: named ratios and angles replace vague impressions with specific values that can be examined individually.
  • A visible basis for each result: landmark review and metric definitions show which geometry produced a finding.
  • A meaningful breakdown: users can see stronger metrics, lower-scoring metrics, and how each part contributed to the summary.
  • More controlled comparison: analyses taken with similar pose, expression, lighting, and camera conditions can help show whether a photo-derived metric changed.
  • More precise questions: a named measurement can help a user describe a specific concern to an appropriate qualified professional without claiming a diagnosis.

URATEN intentionally focuses on what suitable two-dimensional photographs can measure well. That approach keeps the experience accessible, inspectable, and privacy-conscious. It does not attempt to replace the information available from movement, expression, voice, styling, hair, skin quality, three-dimensional anatomy, personality, cultural context, or individual preference.

These boundaries help users interpret the report correctly. The analysis is strongest as a detailed map of visible proportions and model-based relationships, not as a definition of a person's appearance, attractiveness, or worth.

8. Flags, Potential Score, and professional guidance

Prioritized findings

The current engine prioritizes metric scores below 6 out of 10 for closer review and can flag mapped metrics at or below 4 out of 10 as higher-priority findings. Priority reflects the size of the deviation, the metric's weight, and the model's difficulty classification.

We want unusual measurements to be informative, never falsely clinical or needlessly alarming. A flag means that the photo-derived value differs meaningfully from the current scoring reference. It does not mean that the user has a disease, deformity, or medical condition.

Potential Score

Potential Score is a model-based estimate of what the composite could be if lower-scoring mapped metrics moved toward a metric score of 9. Each estimated contribution follows the same fixed weights and is reduced for changes classified as more difficult. The total is capped at 10.

This feature helps illustrate which metrics have more influence within the current model. It does not predict the result of a specific treatment, establish candidacy, account for medical risk, simulate healing, or guarantee that a person's appearance or score will change by that amount.

When professional evaluation is appropriate

URATEN may highlight a measurement that appears unusual or potentially relevant, but it does not diagnose conditions or recommend treatments based solely on a user's photos. Photographs and landmarks cannot establish medical history, symptoms, bite or airway function, skeletal relationships, tissue quality, contraindications, or procedural risk.

When a specific concern could warrant professional evaluation, URATEN explicitly tells the user which reported measurement or concern to discuss and identifies an appropriate type of qualified professional. That professional can examine the person, request any necessary records or imaging, and decide whether the finding has clinical significance. This makes URATEN's output a clearer starting point for an informed conversation without presenting it as medical advice.

9. The role of language models

A language model does not inspect the user's photo, calculate facial measurements, assign metric scores, set the overall attractiveness score, rank findings, calculate Potential Score, or choose protocol actions. Facial geometry, scoring, and protocol-action selection are deterministic and controlled by URATEN's scoring system.

Generative AI is used only in the paid consultant experience. It receives already-calculated results, available protocol context, and information the user chooses to provide in the conversation. It does not receive the user's photographs or raw landmark coordinates.

The consultant's role is to explain existing findings in clearer language and help the user prepare focused questions. It cannot alter the underlying analysis, and it must not diagnose, prescribe, confirm candidacy, or invent a treatment outside the available protocol actions.

10. Our commitment to accuracy and continuous research

URATEN's rating system is continuously researched and improved. We truly care about accuracy, so we treat refinement as a continuing responsibility. We review metric definitions, reference points, accepted ranges, tolerances, weights, quality rules, and presentation as the system develops.

A future scoring version may interpret the same photo-derived geometry differently. This does not remove the value of the underlying measurements. Keeping measurement values separate from model interpretation allows the reported geometry to remain understandable while scoring methods improve.

The current rating model is a structured analytical framework, not a claim that attractiveness has one scientifically settled or universally valid definition. Individual and cultural preferences are real, and no responsible facial analysis should pretend otherwise. URATEN's value comes from applying a clear measurement protocol consistently and showing the user the details behind the result.

How to interpret accuracy and limitations

  • Capture quality sets the foundation. Results are strongest with a straight, clear, evenly lit photograph and a neutral expression. URATEN rejects unsuitable poses to prevent avoidable measurement error.
  • Landmark review improves the input. Users can inspect important frontal points and place the defined profile points rather than relying on an invisible automated decision.
  • Two-dimensional analysis has a clear purpose. It provides accessible ratios, angles, and relative positions from photographs. Questions that require internal or three-dimensional anatomy belong with a qualified professional and appropriate clinical imaging.
  • Reference ranges organize comparison. They make consistent scoring possible, while personal preferences, age-related change, population differences, and cultural context remain broader than any single model.
  • Small changes deserve careful interpretation. Comparisons are most meaningful when capture conditions are closely matched, because a small difference may reflect the photograph rather than a physical change.
  • Health questions stay with professionals. No score or flag establishes pathology, treatment need, candidacy, or likely treatment outcome.

These boundaries are part of responsible analysis. They define what URATEN can do well: turn user-reviewed landmarks into transparent geometric measurements, apply one scoring model consistently, show how each metric contributes, and help users ask more precise questions. The overall score is one summary within that larger body of useful information, never a judgment of a person's value.

Choosing a facial-analysis service

No matter which facial-analysis service you choose, we strongly encourage you to read both its privacy terms and its methodology before providing a photograph. Look for direct answers to important questions: Does the raw image leave your device? Is it sent to a language model or another third-party AI system? Can it be retained, shared, sold, or used to train AI models? Is the result derived from disclosed measurements, or from an opaque visual judgment? A service worthy of your trust should make these answers easy to find.

Your face is personal, and you deserve to make that choice with clear information. URATEN's approach is to keep raw photos on your device, use reviewed landmarks and transparent geometry for scoring, and explain the system's limits openly. Whatever you decide, we wholeheartedly wish you a safe, respectful, and constructive facial-analysis journey.

For more information about how analysis data is handled, read the Privacy Policy. The Terms of Service explains the conditions that apply when using URATEN.