FaceShapeAnalyzer is a free, independent tool that estimates a general face shape category — oval, round, square, heart, triangle, diamond, or oblong — from a single photo, using AI facial-landmark detection that runs entirely in your browser. No account, no photo upload to a server, no cost.
You provide a photo (taken with your camera or uploaded from your device), and the tool runs Google's MediaPipe FaceLandmarker model locally on your device to find 478 facial landmark points: the edges of your forehead, cheekbones, jawline and chin, among many others. From ten of those points it computes four ratios — face length versus your face's widest measured width, jaw width versus cheekbone width, forehead width versus cheekbone width, and how much your jaw narrows toward your chin — and compares those ratios against fixed thresholds to pick the closest-matching shape category. Every landmark index and threshold is listed below; nothing is a black box.
Landmark indices refer to MediaPipe's standard 478-point face mesh: forehead top (10) to chin (152) for face length; cheekbone-to-cheekbone (234–454) for cheek width; mid-jaw (172–397) for jaw width; jaw near the chin (176–400) for jaw taper; and temple-to-temple (54–284) for forehead width. The classifier checks, in order: forehead notably wider than jaw → heart; jaw notably wider than forehead → triangle; cheekbones clearly the widest of the three → diamond; face much longer than wide → oblong; face notably longer than wide → oval; jaw stays close to cheekbone width near the chin (angular) → square; otherwise → round. These thresholds were chosen to produce sensible-looking results across a range of test photos, not derived from a peer-reviewed study — see "What it is not" below.
Anyone curious about their own face shape for choosing glasses frames, hairstyles, or makeup contouring guides; anyone who wants a quick, private, free alternative to uploading a selfie to an unfamiliar app; and anyone who wants to see facial-landmark detection technology in action with a concrete, personal example. It is explicitly not built for, marketed as, or validated as a clinical, cosmetic-industry, or optical-fitting tool — see the next section.
Every photo you analyze is processed entirely on your own device — nothing is uploaded to, or stored on, our servers, because there is no server-side image processing at all. The photo lives only in a JavaScript variable and an on-page canvas for as long as your tab is open, and is discarded the moment you upload a new photo, close the tab, or refresh the page. See the full privacy policy for the complete picture, including the third-party network requests the page does make (loading the face-detection code and model files, which never include your photo).
Questions, corrections, or feedback: support@faceshapeanalyzer.com.