Audio detection
Audio detection listens to a camera's audio for specific sounds, such as glass breaking, an alarm, or a dog barking. It runs alongside video detection and can trigger events and notifications on its own. The camera needs to provide an audio stream.
Setup
Enable the audio plugin for a camera in Settings → Plugins (see Set up sensors). The plugin has no settings of its own: it listens for the sounds selected on the camera and takes its confidence from there.
The sounds it recognizes are glass breaking, screaming, gunshots, alarms (smoke, car, siren), a doorbell, knocking, a dog bark, a cat, a baby crying, and speech.
Settings on the camera
Audio detection is tuned in Settings → Detection, per camera:
- Sounds. Which sounds count for events and automations. All are selected by default. Other sounds covers everything outside the list, including the plain audio alarm some cameras report. With nothing selected, no sound counts.
- Min. volume (dBFS). Audio frames below this level are skipped. Ranges from -100 (silence) to 0 (maximum), default -40.
- Audio confidence. Sounds the detector is less sure about than this are dropped (0 to 1, default 0.7).
- Audio timeout. How long an audio event stays active after the last trigger (minimum 10 seconds).