DEV Community

EmoPulse
EmoPulse

Posted on • Originally published at emo.city

I built an AI that reads 47 biometric signals from your webcam — try it free

During a real investor call last week, my system detected something the transcript tool completely missed: the investor suppressed a positive reaction at timestamp 1:50 — a micro-expression (AU12=0.56) lasting less than half a second.

Circleback, the transcript tool, scored the moment as neutral. My system saw genuine interest hiding behind professional composure.

This is what EmoPulse does. And now it's live for anyone to try.

What is EmoCity?

EmoCity is a real-time biometric scanner that runs entirely in your browser. Point your webcam at yourself and it extracts 47 physiological and cognitive signals in real time:

  • Heart rate from skin color changes (remote photoplethysmography)
  • Micro-expressions — involuntary facial movements under 500ms
  • Stress level — composite index from HRV, facial AUs, cognitive load
  • Authenticity score — Duchenne smile detection (genuine vs. posed)
  • Cognitive load — attention, focus, fatigue indicators
  • Deception proxy — cross-signal coherence checks

All from a standard RGB webcam. All on-device. Nothing leaves your browser.

Try it now

emo.city — open in Chrome, allow camera, start scanning. 3 free scans, no signup.

How it works

Built on Google MediaPipe face mesh (478 landmarks, 52 blendshapes) plus:

  1. FACS-based action unit intensity estimation
  2. Temporal micro-expression detection (sub-500ms)
  3. rPPG heart rate from skin reflectance
  4. Multi-signal fusion into composite indices
  5. Liveness scoring against spoof attacks

Every signal backed by peer-reviewed research: Verkruysse 2008 (rPPG), Wang 2017 (POS), Kim 2018 (HRV+stress), Ekman 1978 (FACS), Yan 2013 (micro-expressions).

Privacy by architecture

All inference runs in browser via WebGL/WASM. Zero network calls during analysis. Open DevTools Network tab during scan — zero outbound requests. GDPR/HIPAA compatible by architecture, not policy.

The backstory

Solo founder, former professional off-road racer (LOORRS Championship), self-taught ML. Entire stack built with 20+ AI agents on Oracle ARM at $0/month. 3 EU patents pending (80 claims).

Try it: emo.city
Architecture: emopulse.app/under-the-hood.html

Built by Arvydas Pakalniskis | EmoPulse | 3 pending patents, 80 claims

Top comments (0)