See where your content loses people and why

Facial expression recognition and behavioural analytics that tell you exactly which section of your training is failing, and what disengagement looks like before someone quits.

engagement_stream.live
--
Engaged
baseline calibratingframe 0
The Problem

Your training is a black box

Completion rates tell you someone finished. They tell you nothing about whether they were actually engaged, where they checked out, or why.

📋

Completion rates lie

Someone can complete a training module while mentally checked out for 80% of it. You'd never know.

Surveys are lagging

Post-session surveys capture what people remember feeling, not what actually happened moment to moment.

🔍

Zero content-level insight

"Learners are disengaged" is useless. You need to know which slide, which segment, and what the emotional pattern looked like before they quit.

Who It's For

Built for anyone delivering content at scale

If you're producing training, learning, or informational content and you need to know whether it's actually working, Emotuit gives you the answer.

🏛

Institutional training

Government agencies, international organisations, and NGOs delivering training programmes to large, distributed teams.

🏢

Corporate L&D

Learning and development teams that need to prove training ROI and identify which content actually moves the needle.

🎓

Education providers

Online course creators, universities, and e-learning platforms optimising content for genuine understanding, not just completion.

How It Works

From webcam to actionable intelligence

A continuous pipeline that runs throughout every session. No hardware. No installation. Just a browser and a webcam.

01
Detect
Face detection and 68-point landmark extraction via MediaPipe
02
Classify
7-emotion probability vector using Ekman's FACS framework
03
Calibrate
Per-user baseline locks within seconds - all readings as deviation
04
Track
Tab-switches, face-loss, window state, focus duration in parallel
05
Score
Continuous engagement index (0-1) from fused emotion + behaviour
06
Correlate
Engagement mapped to exact content timestamps - per slide, per segment
07
Predict
Disengagement model triggers alerts before the learner actually leaves
Research

Every component peer-reviewed

Originally designed in 2014. Every architectural decision independently validated by published academic research.

Baseline Calibration
Confirmed as best practice for affective computing. "Calibration involves capturing a neutral expression and adjusting subsequent data accordingly."
Frontiers in Psychology, Jan 2026
Tab-Switch Tracking
Single most important predictor of disengagement in online learning, outperforming self-regulation and satisfaction.
ScienceDirect, 2024
Multi-Signal Fusion
Combining facial expression with behavioural signals improved classification accuracy from 91.5% to 94.6%.
OUCI/DNTB Online Learning Study
Predictive Labelling
73.3% disengagement prediction accuracy with 40% of the session still remaining.
Boote, Agarwal & Mostow, 2021
Content Correlation
Real-time framework mapping facial engagement to content positions for per-section analysis.
PMC MOOC Framework, 2021
Emotion Classification
Deep learning models using identical 7-emotion engagement indices to classify learner states.
PMC / Springer, 2022
Privacy

No facial data leaves the device

All detection and classification runs client-side in the browser. Only numerical scores are transmitted. Never images. Never video.

Client-side inference

TensorFlow.js and ONNX Runtime Web. All processing in the user's browser.

No images transmitted

Only emotion vectors and engagement scores leave the device. Never pixels.

Explicit consent

Webcam requires clear opt-in. No background activation. No exceptions.

Fully anonymisable

Content-level analytics work without any PII whatsoever.

Air-gap compatible

Zero external dependencies. Runs fully offline and on-premise.

Compliance ready

GDPR, institutional governance, governmental data handling. From the ground up.

Integration

Works with what you already use

A lightweight JavaScript SDK that drops into any existing platform. No migration. No infrastructure changes.

Any LMS
SCORM
xAPI / Tin Can
Custom platforms
Video players
Webinar tools
Browser-based content
Get Started

See what completion rates can't tell you

Find out exactly where your content works, where it fails, and what disengagement looks like before people leave.

Request a Demo