Upload a photo and discover exactly what emotion your face is expressing — with a full confidence breakdown across all 7 emotions, completely free.
For best results: face the camera, ensure good lighting, and use a natural expression.
or drag & drop · JPG, PNG, WebP
Three steps from pixel values to psychological affect classification.
The convolutional neural net scans 68 landmark coordinates across the brows, eyes, and mouth.
It calculates muscle contraction vectors based on the Facial Action Coding System (FACS).
A softmax layer generates a 100% probability curve across all 7 universal emotions.
Micro-expressions convey subconscious feelings faster than words.
Test whether your portrait conveys the intended emotion or intensity.
Choose profile pictures that project approachable, genuine positivity.
Runs locally in your browser via WebAssembly without uploading images to cloud servers.
Radiating joy, warmth, and positivity. The AI detected raised cheek muscles and smile contours around the lips.
Reflective, downcast, or melancholic. Detected through downward lip corners and tension in the inner brow.
Intense, assertive, or frustrated. Detected through furrowed brows, tightened eyelids, and narrowed lips.
Wonder, shock, or sudden realization. Indicated by widened eyes, raised arches, and an open jaw.
Alert, apprehensive, or startled. Marked by pulled-back lips and raised, drawn-together brows.
Aversion or distaste. Indicated by a wrinkled nose bridge, raised upper lip, and squinted eyes.
Calm, composed, and objective. Relaxed facial musculature with no dominant emotional tension.
Human facial expressions are produced by coordinated contractions of subcutaneous muscles around the eyebrows, eyelids, nose bridge, cheeks, and lips — documented in psychology as the Facial Action Coding System (FACS). Our AI Emotion Detector runs a two-stage neural pipeline inside your browser using face-api.js (TensorFlow.js with WebGL/WebAssembly): first, SSD MobileNet v1 locates the face and maps 68 facial landmarks to align the eyes and mouth horizontally; second, a depthwise separable convolutional expression network evaluates muscle tension patterns across the aligned face crop.
Rather than outputting a single rigid label, the model computes a softmax probability distribution across all 7 universal emotions: Happy (cheek raiser + lip corner puller), Sad (inner brow raiser + lip corner depressor), Angry (brow lowerer + lid tightener), Surprised (outer brow raiser + jaw drop), Fearful (brow raiser + upper lid raiser + lip stretcher), Disgusted (nose wrinkler + upper lip raiser), and Neutral (relaxed baseline musculature). This lets you inspect both your dominant expression and subtle secondary undertones in any portrait.
Computer vision reads outward facial muscle geometry in a single still frame, not a person's internal emotional state or psychological mood. Resting facial structure, cultural expression norms, camera angle, and lighting contrast can cause a relaxed face to register mild Sad or Angry undertones. All results are provided strictly for informational entertainment, photography feedback, and self-discovery — not psychological or clinical assessment.
Every expression scan runs locally on your device using WebAssembly and WebGL. Verify it yourself: open your browser DevTools (F12), open the Network tab, upload a photo, and see that no image request is ever sent to any server.
Everything you need to know about facial emotion detection.
