Loading page
About this tool
Analyze your smile with AI-powered facial expression detection. Get smile score, emotion analysis, confidence meter, and detailed report.
The Smile Detector runs a face-detection model with 68-point landmarks and a seven-emotion expression classifier on a photo you supply, then turns the results into a smile score, an expression label and a breakdown of mouth curve, teeth visibility, eye squint, cheek raise, lip symmetry and face symmetry. The smile score is simply the model's happy probability expressed as a percentage, and the headline overall score is a weighted blend that gives 30% to that smile score, 20% to a natural-smile figure, 15% to mouth curve, 10% each to teeth visibility, eye squint and lip symmetry, and 5% to face symmetry. It is a light, for-fun read on a portrait, not a clinical or diagnostic assessment.
Open Smile Detector on AltFTool — it loads instantly in your browser.
Drop a JPG, PNG, WEBP or HEIC photo up to 20MB on the uploader, or use the Paste or Camera option.
The face model places its 68 landmarks and returns an expression label with a confidence figure and the Overall Score, Happiness and Natural Smile gauges.
Read the Mouth Curve, Teeth Visibility, Eye Squint, Cheek Raise, Lip Symmetry and Face Symmetry bars, then press Download Report for smile-report.txt.
Six landmark-derived measures are reported alongside the score, so you can see whether a low result came from the mouth, the eyes or asymmetry.
The confidence figure is the strongest of the seven emotion probabilities, so you can tell a certain neutral from an uncertain guess.
The detection models load into the page and run on your device, so the image is analysed without being uploaded anywhere.
The smile score is the model's happy probability multiplied by 100, so a happy probability of 0.72 gives a smile score of 72. The separate overall score blends that with mouth curve, teeth visibility, eye squint, lip symmetry and face symmetry using fixed weights, with the smile score itself carrying the largest share at 30%.
They are thresholds on the happy probability: above 0.85 is a Big Smile, above 0.65 is Smiling, and above 0.5 is a Slight Smile. Below 0.5 the label falls through to Sad, Surprised, Angry or Neutral depending on which of those probabilities clears 0.4.
Not reliably — treat the fake-smile figure as entertainment. It is derived from where the happy probability sits between 0.5 and 0.7, plus an eye-aperture ratio used as a rough stand-in for the eye crinkling associated with a genuine smile, and neither is a validated measure of sincerity.
Usually because the face is too small, too dark, turned away or partly covered — the landmark model needs a reasonably front-facing, well-lit face to place its 68 points. Use a photo where the face fills a decent part of the frame and both eyes are visible, and re-run it.