I transformed my existing security cameras into an automatic bird identification system using BirdNet-Go, which runs locally via Docker on hardware like a Raspberry Pi. The system uses the cameras' microphones to detect and identify bird songs in real-time, 24/7, without cloud dependencies or subscription fees. It supports 14,795 species with the new Google Perch v2 model, up from 6,000 in BirdNET 2.4, and also detects bats and frogs.

Key features include customizable alert rules for specific species, novelty tracking for first-time yard visitors, and RTSP stream support for standard IP cameras. I connected three existing cameras, turning them into dual-purpose security and biodiversity tools. The system integrates with Bird Weather for community data sharing and Home Assistant via MQTT for smart home notifications — I even pipe alerts to a dedicated Discord channel.

Visual audio energy levels help optimize camera placement by revealing noise interference. The entire stack is self-hosted, ensuring privacy and permanence. As a humorous aside, the system once detected a "fart" in the driveway from a passing neighbor, proving its sensitivity to unexpected audio events.