In the Field

Testing Out Song Sleuth, a New App That Identifies Birds by Their Calls

Song Sleuth's advanced software can recognize the voices of 200 common bird species. But how does it fare in the field?

Last week North American birders got another exciting ID tool to add to their arsenals: Song Sleuth, a new app that’s available in the iTunes store for $9.99 (with a Droid version slated to release this fall). The technology—12 years in the making—combines a sophisticated sound-recognition algorithm with species illustrations, descriptions, and range maps by David Sibley.

Song Sleuth isn’t the first birdsong-recognition app to hit the market. Warblr and ChirpOMatic both came out in 2015, but they mainly cater to a U.K. audience. Instead, Song Sleuth is the first to have an extensive library of North American species. Its list includes nearly 200 birds, with potential future additions, according to Wildlife Acoustics, the Boston-based company behind the technology. The app is self-contained, capable of recording, recognizing, and identifying songs in one smooth, seconds-long process. It also generates spectrograms—graphs that show the spectrum of frequencies in a bird’s vocalization—to help the app analyze recordings better, and allow users to study, learn, and compare the species’ sounds.  

Audubon was asked to test out Song Sleuth before its official launch, so I met with Wildlife Acoustics’ Director of Product Management Sherwood Snyder for a tutorial in Boulder, Colorado. A mechanical and electrical engineer by training, Snyder previously worked at the Cornell Lab of Ornithology’s Bioacoustics Research Program, where he designed hardware and software to monitor whales and other animals.

For all the science behind the app, I found it to be straightforward and convenient, even for a beginner birder like me. I immediately liked the fact that I didn’t need Wi-Fi or data to use it. All the material, including the Sibley references, is available for download, so you can use Song Sleuth in remote locations where getting a signal might be an issue—say, deep in the Rocky Mountain backcountry. In total, the app takes up 367 megabytes of space: less than Facebook, but more than Pokémon GO or Snapchat.

Right after you open the app, you're asked to enter your location. Once I did that, it winnowed down the bird catalog to candidates that’d be present in Colorado in early February (it automatically reads the date). The location also helps the app filter choices by a species’ regional song “dialect.” As soon as you’re set up, it’s time to eavesdrop on a bird. I found the loudest ball of feathers in my vicinity and selected “Record and ID” from the menu icon in the upper left corner of the app screen.

The trickiest part proved to be recording a quality sound byte. It was quite windy for our test, so Snyder advised me to use my back as a shield and hold the phone upside down, aiming the microphone toward the bird I was trying to identify. Ideally, you should record the entire song, but that can be a challenge. It took me a couple tries just to get a clip without too much background noise or other bird sounds. The app then gave me the three most likely matches based on the audio and spectrogram, in order of highest probability: American Robin, White-breasted Nuthatch, and Bald Eagle.

From left to right: Sonograms comparing the field recording with the app's pre-recorded clips; the top three options, ranked by probability. Courtesy of Jayme Moye

After comparing my clips to Song Sleuth’s pre-recorded examples, I was able to verify that I was indeed looking at an American Robin. The app had nailed the ID. If I wanted more input, I could also compare the spectrograms or check the Sibley data, which includes more traditional mnemonic devices (like cheerily, cheer up, cheerily, cheerio for the robin).

But how would the app test on a less conventional call? For fun, I recorded myself doing a bird impression (something between a dying Black-capped Chickadee and a Mourning Dove). The three best matches were: human, Rock Pigeon, and Great Horned Owl. OK, it got me. In addition to human, the app comes with a few other non-avian sounds commonly mistaken for birds, including three types of frogs and toads, plus three types of squirrels.

Another cool Song Sleuth feature is the geotagging functionality. Every time you make a recording, the app marks your location and displays the recording on a map, allowing the technology to refine its selections the next time you revisit the spot. You can also share your recordings with others via text or email—an important feature for individuals who are used to passing their eBird checklists around. It’s also a great way to get the word out about rarities or solicit a second opinion when you think the species is too bizarre for the app’s capabilities.

After meeting with Snyder, I was excited to try the app out on other field excursions. But I also made a note to use it to hone my birding skills at home. I now have a convenient catalog of birds, tailored to my city, that I can cruise through while chilling on my couch. Maybe it’s time to cancel that Spotify subscription and finally figure out all of my sparrow songs. 

“The views expressed in user comments do not reflect the views of Audubon. Audubon does not participate in political campaigns, nor do we support or oppose candidates.”