One bird recording with every call placed in three dimensions, similar-sounding calls grouped together

Syrinx 3D

Hear a bird.
See its song.

Record a bird singing and Syrinx tells you which bird it is. Then it takes the recording apart, call by call, and shows you the shape of what it just sang.

Free. No account. Works on a phone.

Every other app stops at the name.

Knowing it was a robin is the easy half, and it is where the other apps put their full stop. Syrinx keeps going: it separates the song into the individual calls inside it, measures each one, and places it so that calls which sound alike end up near each other. What you get back is the structure of the song, its repeats and its phrases, as something you can look at rather than only hear.

1

Record it

Ten seconds on a phone is enough. In the browser, or the Android app. Nothing is uploaded.

2

It names the bird

BirdNET over overlapping windows, with the confidence shown rather than hidden. Calls too short to name alone can be named from the group they belong to.

3

It draws the song

The recording is cut into its individual calls, each one measured, and laid out so calls that sound alike sit together. Tap any one to hear it on its own.

Free
No account, no trial, no paid tier.
On your device
The analysis runs locally. Your recordings stay yours.
197 species
Real field recordings bundled in, to try before you go outside.
Android too
If you would rather record without a browser in the way.

We are building a dictionary
of what birds are saying.

Science has pinned down a handful of calls. A chickadee adds more dees to its alarm when the predator is more dangerous, and the birds around it react accordingly. Past a few examples like that, the meaning of almost every bird sound on earth is simply unrecorded.

That is not a hard audio problem. Machines got good at hearing birds years ago. It is a missing dataset: nobody has paired this exact sound with this is what the bird was doing at anything like the scale it would take.

That is the part you can do

When you record something, write down what you saw. Chasing off a rival. Warning about a cat. Calling a mate. Feeding young. Nothing in particular, which counts too. It takes seconds and it is the half a model cannot supply: a machine can tell you a wren made that sound, but not that there was a cat under the feeder at the time.

Enough of those line up and a sound starts to carry a probable meaning instead of only a species name. Every label is one row in a dictionary that does not exist yet.

Where it honestly stands: early. Enough labels to see structure, not enough to make confident claims, and the app says so on screen rather than dressing up a guess. It may also not resolve the way people picture it. Birds may not have words, and some of what sounds like language may turn out to be closer to tone of voice. We would rather find that out than pretend otherwise.

The longer version

There is a bird outside right now.

Open the analyzer

Made by one person. How to use it · Bird song guides · Contact