Live
Live stream offline — showing latest snapshot.
Sentinel listens around the clock and runs bird-sound inference on the device. Its public page shows only high-confidence candidate labels with source audio for human review. Other wildlife models stay private until they pass validation. Deploy it anywhere, on sun and satellite.
Bird candidate scoring runs on the camera itself. There is no per-detection cloud charge, and processing continues through a thin or intermittent link.
Selected high-confidence candidate labels can appear on a no-login page with playable source clips, model scores, weather, and an explicit warning that AI labels are not verified observations.
Solar power and a satellite uplink mean you can drop a station in a marsh, a canyon, or a roadless preserve, with no grid and no cell signal, and watch the species roll in from anywhere.
These are the actual public pages from a Sentinel camera at the Alice Ferguson Foundation in Accokeek, Maryland. BirdNET proposed each public label on the device; the labels remain candidates until a qualified reviewer verifies the audio.
BirdNET is the only detector currently allowed into the public feed, limited to bird candidate labels scoring at least 90%. The historical BirdNET archive is included at that same score floor and remains explicitly labeled as unverified AI output. Frog, insect, mammal, and reptile models are private research tracks; none are presented as live identifications. Google Perch v2 runs only in an isolated private validation process and is not used as a public species authority.
A reporting assistant can summarize candidate-label frequency and daily rhythm without turning model output into a confirmed presence or biodiversity claim. Verified conclusions require human review and an appropriate sampling design.
We deploy and manage Sentinel Wildlife stations for land trusts, preserves, parks, and research sites. Send a note and we'll talk through fit and pricing.