FaceShape Lab

Face Emotion Detector - AI Smile & Expression Analyzer

Detect emotions, smile intensity, age, and gender from a photo or live camera with our free AI Face Emotion Detector. Instant results with facial blendshapes.

Expression scan

Upload a clear front-facing face photo.

Expression results appear after scanning

Smile intensity via facial blendshapes, emotion recognition (Happy, Sad, Angry, Surprised, Fear, Disgust, Neutral), plus AI age and gender estimates. Use live mode for continuous updates from your camera.

Why Use an AI Face Emotion Detector?

The Face Emotion Detector reads smile intensity from 52 facial blendshapes and classifies emotions with an AffectNet-trained neural network, alongside AI age and gender estimates — from a single photo or your live camera.

1

Smile Detection via Blendshapes

MediaPipe facial blendshapes measure mouthSmileLeft and mouthSmileRight independently, so asymmetric smiles are captured precisely.

2

7+ Emotion Recognition

Happy, Sad, Angry, Surprised, Fear, Disgust, Neutral, and Contempt are classified with an AffectNet-trained EfficientNet model.

3

Live Camera Mode

Go live and the analyzer updates every second from your webcam using fast on-server models — no video is ever stored.

4

Ensemble Age Estimate

Photo uploads run the full MiVOLO + FairFace age ensemble; live mode uses a fast FairFace pass for instant updates.

5

Gender Detection

InsightFace's production genderage model estimates Male / Female with a detection confidence score.

6

Privacy First

Frames are analyzed in memory and discarded; expression analysis never saves your photo or camera stream.

How Does the Face Emotion Detector Work?

Three steps — or go live and let it update continuously.

1
📷

Upload or Go Live

Upload a front-facing photo, or start live mode to stream analysis from your camera every second.

2
🧑‍💻

AI Analysis

Facial blendshapes measure your smile while neural networks classify emotion and estimate age and gender.

3
📊

Result

See your dominant emotion, smile meter, full emotion distribution, and age/gender estimates instantly.

What the Expression Analyzer Measures

Every scan reads independent facial signals and reports them separately, so you can see exactly what drives the result.

Smile Intensity (0-100)

Left and right mouth-corner blendshape coefficients, averaged into one smile score with a per-side breakdown.

Emotion Distribution

Probability scores across all eight emotions, not just the winner, so borderline expressions stay transparent.

Apparent Age Range

A point estimate with an honest low-high range from the two-model ensemble rather than fake single-number precision.

Gender Confidence

Male / Female classification with the detector's confidence score attached.

⚖️

Model Standards

Production models only — no hand-tuned guesswork.

1Smile: MediaPipe FaceLandmarker blendshapes (52 coefficients)
2Emotion: EfficientNet-B0 trained on AffectNet (8 classes)
3Age: MiVOLO + FairFace ViT ensemble with outlier rejection
4Gender: InsightFace buffalo_l genderage head
5Live mode budget: under one second per frame on CPU
🔒

Privacy by Design

Photos and live frames are processed in server memory and discarded immediately. Nothing from the expression analyzer is stored, logged with your identity, or used for training.

Private

No public image sharing

Secure

Browser-safe analysis flow

Face Emotion Detector FAQs

1

Which emotions can the detector recognize?

+

Happy, Sad, Angry, Surprised, Fear, Disgust, Neutral, and Contempt — classified by an EfficientNet model trained on the AffectNet dataset, which covers real-world photos far better than older FER2013-era models.

2

How does smile detection work?

+

We use MediaPipe facial blendshapes — the same technology behind AR face filters. The mouthSmileLeft and mouthSmileRight coefficients are measured separately and combined into a 0-100 smile score.

3

Is my camera stream stored anywhere?

+

No. In live mode a single frame is sent roughly every second, analyzed in server memory, and immediately discarded. Nothing is written to disk or any database.

4

Why does live mode show a slightly different age than photo upload?

+

Live mode uses a single fast model to keep updates around half a second, while photo uploads run the full two-model ensemble for the most accurate estimate.

5

Can it detect emotions on multiple people?

+

The analyzer focuses on the largest, clearest face in the frame for the most reliable reading.

What Users Say

D
★★★★★

“The live mode is addictive — smile meter reacts instantly and the emotion bars update as my expression changes. Impressive that nothing gets stored.”

Danish R.

Dubai

E
★★★★★

“Tried it for my presentation practice. Watching the neutral-to-happy shift in real time helped me calibrate a more approachable expression.”

Elena M.

Berlin

K
★★★★★

“Uploaded a family photo and the emotion breakdown matched everyone's mood perfectly. The age estimate with the range felt honest rather than overconfident.”

Kevin O.

Toronto

D
★★★★★

“The live mode is addictive — smile meter reacts instantly and the emotion bars update as my expression changes. Impressive that nothing gets stored.”

Danish R.

Dubai

E
★★★★★

“Tried it for my presentation practice. Watching the neutral-to-happy shift in real time helped me calibrate a more approachable expression.”

Elena M.

Berlin

K
★★★★★

“Uploaded a family photo and the emotion breakdown matched everyone's mood perfectly. The age estimate with the range felt honest rather than overconfident.”

Kevin O.

Toronto

D
★★★★★

“The live mode is addictive — smile meter reacts instantly and the emotion bars update as my expression changes. Impressive that nothing gets stored.”

Danish R.

Dubai

E
★★★★★

“Tried it for my presentation practice. Watching the neutral-to-happy shift in real time helped me calibrate a more approachable expression.”

Elena M.

Berlin

K
★★★★★

“Uploaded a family photo and the emotion breakdown matched everyone's mood perfectly. The age estimate with the range felt honest rather than overconfident.”

Kevin O.

Toronto