How It Works
AuraTwin reads your facial expressions in real time, maps them to emotional states, and turns that data into two metrics: a wellbeing score and a burnout risk indicator.
Emotion Detection
A camera captures your face. A face detector finds it in the frame, then the emotion model classifies what it sees into one of 8 categories.
The model: enet_b0_8_best_afew
An EfficientNet-B0 network fine-tuned for facial emotion recognition. EfficientNet-B0 is a compact image classifier โ accurate without requiring heavy hardware. It runs via ONNX Runtime, which keeps memory usage low and inference fast.
av-savchenko/hsemotion-onnxTraining data: AffectNet
~450,000 real-world facial images collected from the internet โ not posed or staged. People photographed in natural conditions across a wide range of demographics, lighting, and angles.
8 Detected Emotions
Wellbeing Score
Each emotion detection maps to a 0โ100 score based on how positive and how calming that emotion is. Your score is the average across all detections in the last 28 days.
Based on Russell's Valence-Arousal model
Emotions sit on two axes: valence (how positive or negative it feels) and arousal (how energizing or calming it is). A relaxed, happy moment scores much higher than an anxious or angry one.
Formula
score =
ย ย (valence ร 0.7 + (1 โ arousal) ร 0.3) ร 100
Valence is weighted at 70% because positivity has a stronger effect on wellbeing than energy level.
Score Ranges
Emotion Values
Burnout Risk
AuraTwin also tracks whether your emotional patterns suggest you might be heading toward burnout โ scored 0โ100%, calculated over the last 28 days.
Emotional Exhaustion
The share of moments where Sad, Angry, or Fear was detected. The more often these appear, the higher the exhaustion score.
Accomplishment Loss
How rarely Happy appeared: 1 โ Happy ratio. If you're almost never detected as happy, this score rises.
Overall Risk
A weighted average of both. Emotional Exhaustion carries more weight โ sustained stress is the strongest early sign of burnout.
Inspired by the Maslach Burnout Inventory
The MBI is a questionnaire tool widely used to measure burnout in workplace settings. AuraTwin adapts its two core dimensions โ exhaustion and reduced accomplishment โ into a metric derived from observed emotion data rather than self-reported answers.
Trend Direction
The 28-day window is split in two. Comparing the halves shows whether things are getting better or worse.
Your Digital Twin
All the data from detection, scoring, and risk analysis feeds into a living representation of your emotional state โ a digital twin that updates as you do.
The Face
Your dashboard shows an animated avatar whose expression reflects your current wellbeing score. It updates in real time as new emotion data comes in โ a quick visual signal that doesn't require reading numbers.
The face also animates differently per state โ bouncing when happy, slow breathing when neutral, wobbling when exhausted.
AI-Powered Reports
On demand, AuraTwin generates a written report summarising your emotional patterns. It reads your emotion history and produces a plain-language analysis โ what you've been feeling, how it's changed, and what to watch.
Powered by Gemini 3.1 Flash Lite
Google's fast multimodal model interprets your data and writes the report.
Based on your actual data
The model sees your emotion logs, wellbeing score, burnout risk, and trend direction.
Generate any time
Reports are stored so you can look back and compare across weeks or months.
What AuraTwin Is โ and Isn't
Honest about what it can and can't do.
It is
- โA tool for seeing your emotional patterns over time
- โA personal awareness aid based on observed facial expression data
- โA complement to โ not a replacement for โ professional support
It is not
- โA medical device or diagnostic tool
- โA substitute for mental health care
- โA perfect or clinical measurement of emotional state