Posture Analysis Tool

Upload a photo and get instant forward head posture, shoulder level and hip level measurements from AI body-keypoint detection — computed entirely in your browser. Nothing is uploaded, stored, or sent anywhere.

100% on-device — photo never leaves your browser Side & front profile Free · no account Powered by MoveNet (TensorFlow.js)

Analyze a photo

JPG, PNG, HEIC or WebP. Best results: plain background, full torso (and head) visible, decent lighting, camera roughly waist-to-chest height.

How this posture checker works

This tool runs TensorFlow.js's MoveNet pose-detection model directly in your browser to find 17 body keypoints in your photo — ears, shoulders, hips, knees and more — then computes simple geometric ratios between them: how far forward your ear sits relative to your shoulder (a side-profile marker sometimes associated with forward head posture), and how level your shoulders and hips are relative to each other (a front-view marker of left/right asymmetry). Full formulas and limitations are on the How It Works page.

Why nothing gets uploaded

The model — a small neural network — is downloaded once to your device, and every photo you analyze after that is processed locally by your browser's own GPU (via WebGL) or CPU (via WebAssembly). There is no server-side image processing on this site at all, which means there's no server log, no storage bucket, and no possibility of your photo being seen by anyone else — a meaningful difference from apps that require uploading photos to a company's servers for "AI analysis."

FAQ

Is this a medical or diagnostic tool?
No. It reports descriptive geometric measurements from your photo's keypoints — nothing more. It cannot diagnose a postural condition, assess pain, or replace an in-person evaluation by a physical therapist, chiropractor, or physician.
Which photo should I use — side or front?
Side profile is more informative for forward head posture (the classic "text neck" pattern). A front-facing photo is better for checking whether your shoulders or hips sit level with each other.
Does my photo get uploaded anywhere?
No. Pose detection runs entirely in your browser using TensorFlow.js. See the privacy policy for the full explanation.
Why did it say "no person detected"?
MoveNet needs a reasonably clear, well-lit, mostly unobstructed view of a person. Busy backgrounds, very tight crops, extreme angles, or low light can all cause detection to fail or be unreliable.
How accurate are the measurements?
They're consistent given the same photo, but they are not lab-grade — camera distance, lens choice, and how squarely you're facing the camera all shift the numbers. Treat results as a rough, repeatable reference point over time (e.g. photos taken the same way a few weeks apart), not an absolute clinical measurement. Details on the methodology page.

Informational tool only — not a medical device and not a substitute for professional evaluation.