Real-time AI-powered infant sound pattern monitoring — edge inference on an ESP32-S3, streamed to a calm parent dashboard.
CrySense AI is an experimental audio pattern recognition system. It does not diagnose medical conditions and cannot determine with certainty why a baby is crying.
Live pipeline
🍼 Likely Hungry
Confidence 82% · end-to-end 440 ms
Live status, confidence and latency updates the moment a sound pattern is detected.
TinyML inference runs on-device, so audio never has to leave the room to be classified.
Test and demo the whole pipeline from a browser before hardware is connected.
Realtime Database sync for device state, events and settings, with per-user access.
Measure the room's noise floor and tune sensitivity in a guided wizard.
Curate datasets, extract MFCC features, train and evaluate an experimental model.
Consent metadata on every sample, export your dataset, delete all cloud data.