“Shazam for bird calls” is a compelling sales pitch, and a difficult technical proposition. It’s much harder for an algorithm to identify animal sounds, in all their wild pied beauty, than to recognize a specific recording of a pop song. Impressively, though, several apps really can identify a bird by its song, with a little human help.
We tested four birdsong recognition apps that claim to detect bird calls and songs (which are two different things). The results were impressive.
Methodology
We tested our apps using these recordings of common American birds: the red-winged blackbird, yellow-rumped warbler, mourning dove, American robin, American crow, house sparrow, and a grey squirrel as a gotcha question. Some recordings isolate the featured bird, while some include background birdsong from other species.
Recognizing recordings isn’t the same as identifying an unknown bird in the wild. Some of these apps base their results on the user’s location, which could throw them off when we test them in a Manhattan office building. This isn’t a comprehensive test of the considered apps, but a rough demo. In the field, your mileage will vary.
Still, if you have a recording that you want to run through multiple apps, you’ll need to save the file and play it out loud—none of these apps support file imports. So an app that can recognize calls through a recording is doubly useful.
Results
Among our four apps, we got at least one correct guess on all seven recordings. Three apps performed well, but on different birds. Here’s a breakdown of correct identifications, with Y for yes, N for no, and K for kinda (when the app needed multiple tries or suggested several other birds first).
Three of our four apps performed fairly well. Once you get a recording of your bird, you might want to run it by all three.
Song Sleuth ($9.99 iOS)
The slickest and most sophisticated app. If you want to learn about the birds you ID, then this app is your best choice.
As soon as Song Sleuth boots up, it starts capturing sound, displaying it on a spectrogram. You don’t have to anticipate the bird call. Instead, hit record shortly after the call starts, then stop when you have enough.
You can record indefinitely, though the app recommends trimming your results as short as possible. Remember to delete long recordings to save space.
Song Sleuth lets you edit your recording to isolate the desired birdsong, which is essential in a noisy environment. The editing process is a bit complicated but still intuitive enough to figure out without a tutorial.
The app then analyzes your recording right on the device, and returns a list of likely options based on your recording and your location, with playable samples to compare to. If the app can’t identify your bird, you can still manually choose from its entire database, which even includes squirrels, chipmunks, humans, and the spring peeper chorus frog. The database also provides more sound samples, plus distribution maps, images, and descriptions.
Once you’ve identified your bird and made notes, Song Sleuth saves it to a map of all your recordings. You can also edit, export, or share your past entries. You can even re-analyze an older recording.
In our test, Song Sleuth was the only app to recognize a squirrel—useful for clueless dilettantes like me who might mistake angry rodents for birds.
ChirpOMatic ($3.99 iOS; £3.99 UK iOS)
The simplest app and the best guesser. If you just need accurate guesses and not much information about the birds, this is your best choice.
ChirpOMatic also opens on a recording screen, but it doesn’t start recording until you tap the big red button. Tap again to stop recording, or let it record the maximum 12 seconds.
The app automatically analyzes your recording, factoring in your location and the current season, and returns a handful of guesses along with a photo and sample recording for each bird. If you have a good phone signal, ChirpOMatic will send your recording to a server for a more thorough instant analysis. If you’re offline, it will still locally analyze the recording, and you can save it for a full analysis later. Once you’ve found the right bird, you can confirm it (and return to it later).
ChirpOMatic doesn’t have much info on the birds; for that, you’ll need the companion app, Chirp!, which has brief offline entries, plus links to each bird’s entry in the Cornell Lab of Ornithology and Wikipedia. You’ll also need Chirp! to check out the full database of birds if ChirpOMatic’s guesses are all incorrect. There’s no map view either.
ChirpOMatic includes a “bird-safe” silent mode to avoid playing recordings out loud and confusing the real birds around you.
In our unscientific test, ChirpOMatic eventually guessed every bird species, though sometimes it took a few tries. It couldn’t recognize anything that’s not a bird, which only matters if you’re IDing on sound alone. ChirpOMatic regularly updates its databases with new birds, so you’ll keep getting your money’s worth.
Bird Song Id ($4.99 iOS/Android; £3.99 UK iOS/UK Android)
An okay backup when you can’t get a good recording. Only use this if the first two apps aren’t covering all your needs.
Bird Song Id, which comes in US and UK flavors, has an ugly but functional interface. It performed slightly worse than our other two options, but it recognized a couple of calls that gave ChirpOMatic trouble.
You’ll have to hunt for the detection function: From the home screen, select “Auto Rec” and hit the recording button. Try to get at least 30 seconds of birdsong. When you’re done recording, hit stop, then “Automatic Recognition.”
The app has a strong results page, which includes a confidence score for each possible bird. You can play samples of the possible birds right from the results.
The “My Recordings” section displays your previous recordings by date or on a map. Again, the interface is ugly but data-rich.
If you can’t get a good recording, as a last-ditch option, try answering the questionnaire in the “Manual Id” section, specifying the pitch, regularity, volume, and region of the call. To process your answers, hit the check mark—not the inexplicably placed “Quiz” button, which autoplays bird calls so you can train yourself to identify them.
SongBird ($9.99 iOS, deleted)
Pigeon droppings. A lesson in researching before buying.
This app, deleted from the App Store days after we bought it, failed to ID or even attempt to ID a single bird call. We tried eleven additional bird recordings and three actual birds outside our window, and we got the same opaque error message every time. The app contained no bird information, just a recording screen. If it wasn’t a scam, it was the worst-built app we’ve ever seen. (Apple refunded our ten bucks after a polite email.)
We include this only as a cautionary tale. Don’t pay for apps without reading reviews and researching for some evidence that they work. Even Apple’s supposedly stringent submission requirements failed to screen out this app. Meanwhile it racked up over 400 upvotes on the consumer forum Product Hunt, even as the developer fought with disgruntled users in the comments.
Conclusions
If you pick just one app to identify bird calls, make it Song Sleuth. Its automatic recording, sophisticated editing interface, and comprehensive bird database make it the strongest all-around tool.
If you have any trouble with Song Sleuth, ChirpOMatic and Bird Song Id can pick up some of the slack. It’s just $14 for all three—not a bad price to feel like Snow White communing with the bluebirds. Detecting bird sounds is a very challenging task for a computer, one that no app can do perfectly. While your brain can tune out all the other birds in the background, it’s much harder for the current generation of algorithms to do the same, and tell which of five birds you meant to identify.
Remember not to play your bird sounds out loud while you’re capturing audio. Birds make these sounds for life-or-death reasons, and if the real birds hear these recordings, it might confuse them and affect their behavior. Don’t be a dick to birds.