FaceGrader — Free AI Facial Geometry and Attractiveness Test

What is FaceGrader?

FaceGrader is a browser-based facial analysis tool that scores a front-facing photo from 0 to 100 based on facial geometry rather than opinion. It detects 478 facial landmarks using MediaPipe's canonical face model and measures four dimensions: symmetry (30 mirrored landmark pairs), facial thirds, facial fifths, and reference ratios. These are combined into a composite score through a formula the site publishes openly — 0.35 × symmetry + 0.25 × thirds + 0.25 × fifths + 0.15 × ratios.

The scoring engine is labeled "v1.0 · FROZEN," which means the same photo always produces the same score. The site is explicit that a high score means closeness to one fixed reference model, and that this is "not a ranking of attractiveness." Its FAQ further clarifies that the reference face is a fixed model derived from MediaPipe coordinates, not a claimed beauty ideal.

How FaceGrader Works

The workflow is a single step: upload a clear, front-facing photo. Landmarks are measured on-device in the browser, and the tool returns a composite score, four sub-scores, and the highest and lowest scoring dimensions. Head rotation, tilt, multiple faces, low resolution, or unreliable geometry cause the photo to be rejected.

Results can be shared as a PNG score card that contains no photo, or through an optional numeric share link. Measurements and results persist only in the local tab session — the site's data receipt claims zero photos uploaded, zero server-stored measurements, and zero accounts, with optional Google Analytics counting visits only.

Main Features

• Geometry-based scoring — a published weighted formula combines symmetry, thirds, fifths, and ratios into a deterministic 0-100 score.
• Six sub-analysis tools — separate tests for symmetry, canthal tilt, jawline analysis, pupil spacing, a "true mirror" view, and the golden ratio.
• On-device processing — photos and measurements never leave the browser, so there is nothing to leak from a server.
• Anonymous score cards — the shareable PNG contains no photo, and numeric share links are optional.
• Frozen engine versioning — v1.0 is fixed, so scores are reproducible and comparable across sessions.
• Documented FAQ — explains the reference model, rejection criteria, and what the score does and does not mean.

Pricing

FaceGrader is free. The site states plainly: "No server upload, no account, no payment." There are no plans, no credits, and no visible paid tier. The only third-party element is optional Google Analytics, which the site says counts visits only.

Common Use Cases

• Casual self-assessment — checking your own facial geometry out of curiosity, with no sign-up friction.
• Understanding specific measurements — the sub-tools isolate single metrics like canthal tilt or the golden ratio that people commonly ask about.
• Social sharing — posting an anonymized score card that shows numbers without exposing the photo.
• Comparing geometry claims — the frozen engine makes it possible to re-test the same photo later and get the same result.

Limitations

FaceGrader measures closeness to one fixed geometric reference model; the site itself warns this is not a general attractiveness ranking, and the score should not be read as a verdict on appearance. Input requirements are strict — head rotation, tilt, multiple faces, and low resolution all cause rejection, and lighting, camera angle, and expression change results. Side-view projection is not measured, so profile analysis is out of scope. The scope ends at still-photo geometry: there is no video, no tracking over time, and no accounts or saved history, since results live only in the open tab. No company, maker, country, or launch information is published on the site, so the operation behind the tool is not identifiable. Finally, like any single-number face score, it invites over-interpretation; the FAQ's own framing is the best guide to what the numbers mean.

Summary

FaceGrader is a small, free, privacy-conscious utility that does one thing: score a front-facing photo against a fixed geometric reference using a published formula. Its strengths are transparency — the formula, landmark counts, and limitations are all documented — and on-device processing that removes upload risk entirely. Its weaknesses are a narrow scope, strict photo requirements, and results that are only as meaningful as a single geometric reference model can be. For casual curiosity about facial symmetry and proportions, it delivers exactly what it documents; for anything beyond that, it is the wrong tool by its own admission.