Capture
Take or upload a front photograph and a profile. Built-in camera guides ensure proper head leveling, distance, and pose.
HarmonyLabs measures facial geometry from front and profile photographs using anatomical landmarks, exact angles, and ratios. Our Harmony score provides an objective reference without generative AI guessing or subjective attractiveness claims.
Step-by-step pipeline
Every metric is grounded in verifiable geometry. You retain complete visibility and control over every landmark proposal.
Take or upload a front photograph and a profile. Built-in camera guides ensure proper head leveling, distance, and pose.
Computer vision proposes landmark coordinates. You can pan, zoom, and drag points that miss your actual anatomy.
Distances, ratios, angles, and facial thirds are calculated directly from verified 2D coordinates without approximation.
The Harmony score maps each measurement against transparent clinical and anthropological reference bands with full formulas exposed.
See how results are presented: hierarchical scoring, category breakdowns, and transparent metric deviations.
Research-informed facial proportional score. Front and profile stay visible separately. Stronger evidence receives greater influence.
Midface ratio: 1.02
Shown against its aesthetic target and harmony range, with the evidence tier beside the number.
Nasofacial angle: 31.2°
Slightly below the typical 34°–38° reference range.
Gonial angle alignment
Accounts for +0.24 potential points on the profile harmony score.
Scientific foundation
Most facial apps apply arbitrary AI filters or black-box neural ratings. HarmonyLabs executes an open, deterministic mathematical pipeline.
Landmarks are localized client-side with a pinned MediaPipe Face Landmarker (float16, task version 1). Coordinates stay on the photograph. Face photos are not sent to a language model.
Side-profile photos are automatically leveled along the Frankfort horizontal plane (porion to orbitale) to compensate for minor head tilts.
Ratios, proportions, and angles are scored against documented plastic surgery and anthropological reference intervals using Gaussian decay curves.
Harmony is an evidence-weighted comparison of facial geometry. Front and profile stay visible separately. Every measurement shows its evidence tier and contribution.
Facial photography is sensitive biometric data. HarmonyLabs never uses your images to train machine learning models.
Landmark detection runs locally in your browser session. Stored images are encrypted, authenticated, and never served via public unauthenticated links.
You can delete any analysis and its associated photographs at any time from your dashboard or settings.