about aviary

There is a microphone in a backyard in Hartsdale, New York. It listens all day and all night. Two models take turns guessing what it just heard: BirdNET-Go, an open source build of the Cornell Lab's BirdNET, and Google's Perch. A Blue Jay at 6:04 in the morning, a Northern Cardinal at dusk, a katydid that will not stop.

There is no feeder here. Hart's Brook Nature Preserve is close by, and a lot of species swing through on their own schedule. The microphone catches them on the way past.

This site is what those guesses look like when you leave them running for a few months. Nobody picked the birds on the page. They picked themselves by showing up. The counts and the times are real, and the recordings are the actual audio the microphone captured.

Some illustrations come from the AvianVisitors bundle and other AI generated bird assets, matched to each species. Not every visitor has a portrait yet.

Aviary started as a fleeting idea and turned into a full end to end experience: a mashup of various open source projects, self-hosted on a machine in the basement. The same pages drive an e-ink display, a Spectra 6 panel running custom firmware fed by a companion backend called pigment, so you can see what is outside without opening a browser.

The site does not use accounts, tracking, or ads. Just birds.