I Put a Microphone in My Backyard and Built a Bird Pokédex
A always-on backyard listener that identifies bird calls, keeps a running journal of who visited, and turns your sightings into a personal field guide — no cloud, no subscription.
The project is on GitHub: github.com/grqg-dev/bird-id.
There’s a microphone pointed at my yard in Santa Barbara. It runs all day. When a bird calls, the software hears it, names the species, saves the clip, and adds another entry to a personal collection I’ve started calling the Bird-Dex — like a field guide, except every page is a bird that actually showed up at my house.
I didn’t build this because I’m a birder. I built it because the question is obvious once you live somewhere with trees: who was that?
What It Actually Does
The idea is simple. Leave it running. Come back later.
The microphone captures short stretches of audio — a few minutes at a time. Each stretch gets analyzed for bird calls. When something matches, the software logs the species, how confident it is, and the exact moment it happened. If a bird was heard, the audio is kept. If the stretch was just wind or silence, it gets thrown away.
Over days and weeks, you end up with a journal of your yard. Which species showed up. How often. What time of day things got loud. And, for the ones you care about, a recording you can play back — the actual call from the actual visit, not a stock sound from the internet.
The Bird-Dex
The part I care about most is the browse view — a web page on your own machine that turns the log into something you’d actually want to scroll through.
Each species gets a card. An illustration, if we have one for that bird. A little waveform picture of the call. A play button. Stats: first time you heard it, last time, how many visits. They’re numbered in discovery order, so species #001 is the first bird that ever registered — your yard’s origin story.
There’s also a gallery mode that’s just the pictures and the names, for when you want to flip through the collection without the audio.
How It Knows What It’s Hearing
The identification comes from BirdNET — the sound-ID model developed by the Cornell Lab of Ornithology and the Chemnitz University of Technology. It’s the same family of tech birders use in apps like Merlin. The model runs entirely on your own computer. Nothing gets uploaded. No API key, no account, no sending your backyard audio to a server in Virginia.
That matters for two reasons. Privacy, obviously — it’s your yard. But also reliability: once the model is loaded, it works offline. The dashboard works offline too. The whole thing is meant to live on a small always-on machine — a Mac mini, eventually a Raspberry Pi — with a USB microphone and no dependency on anyone else’s uptime.
Accuracy improves a lot when you tell the software where you are. A Santa Barbara backyard in late spring has a different cast of characters than a Minnesota lake in October. The location filter keeps the model from suggesting birds that don’t belong in your neighborhood — which cuts down on the embarrassing false positives that come from treating every chirp like an open-world search.
The Illustrations
The cards wouldn’t feel right with blank squares, so each common local species got a small field-guide-style illustration — a mourning dove on a twig, an Anna’s hummingbird in profile, that sort of thing.
I generated them with Google Gemini. You describe what you want — a grid of realistic bird illustrations on a white background, one species per cell, accurate plumage — and Gemini paints the sheet. Then you slice out the individual birds and match them to the species list.
Gemini does not always paint the bird you asked for in the cell you asked for. That part was… an experience. Some illustrations came from a second generated sheet. A few were done one bird at a time when the grid kept giving me a gnatcatcher where I wanted a towhee. The result is good enough that the dashboard feels like a real collection, which was the whole point.
What’s Next
Right now it runs on my Mac. The plan is to move it to something that can stay on 24/7 with a microphone hanging outside — a permanent listener for a permanent question.
I’d like alerts when something new shows up. A species you’ve never logged before. A bird that’s rare for your area. A ping that says: you might want to look outside.
Until then, the microphone stays on, the list keeps growing, and every few days I open the Bird-Dex and find out who came by while I wasn’t paying attention.
The yard has always been full of birds. Now I have names for them.