Which is the best birdsong ID app? We tested them all and have a winner.

     After years of yearning for a technology that can identify birds by their songs, birders are now awash in apps and services that promise to tell you what species you just heard.

     And in fact, the best of more than a half dozen smartphone apps do a pretty good job with the technically complex task of capturing birdsongs and instantly listing the likely birds behind the song.

     But which is best for you? How hard are they to use? And should you go with one of the free versions or spend up to $30 a year?

Leading the field is a free app called BirdNET, which just recently launched an IOS version to go with a popular Android version that has more than a million users.

     We put them all to the test and have clear rankings to share – along with a few surprises.

     Before we delve into how these services perform, it’s worth spending a couple of graphs on the scientific competition that’s behind this progress – and why the science of bird sounds is now coming into its own.

     The bird ID apps are mere sideshows in a race to master the use of sound to study wildlife. The science of bioacoustics, which uses artificial intelligence to analyze audio from the wild, has become a powerful practice in field research.

Click here for our primer on bioacoustics that ran in the Washingon Post.

     The discipline is fueled by the same digital advances powering the internet and has turned out to be particularly effective to study birds. That’s because birds are so vocal and can be found and researched almost everywhere.

A worthy contender is a European app called ChirpOMatic.

     As several major universities, the Cornell Lab of Ornithology and a number of private companies developed bioacoustic techniques, they found they could spin off commercial phone versions that identify bird calls and songs the same way the popular Shazam app does with music.

     About a dozen birdsong applications have reached market as the race to create truly reliable versions has accelerated over the past three or four years.

     Today, there are about a half dozen worthy apps, including versions from the Cornell Lab, Princeton University Press, the technology company Wildlife Acoustics, a couple of European firms and several individual U.S. developers who are betting there’s money to be made on these products. Almost all are available for download for both Android and IOS phones. Other more sophisticated bioacoustic devices that can be used in birders’ yards are also under development. 

     While these solutions are cresting, we couldn’t find anything that ranked how accurate they are. So we spent time with each of them, testing their precision and assessing how easy they are to use. There’s one clear leader in the field, three that were right about half the time and several that were almost always wrong in our tests.

     Here are our findings (and we explain our testing approach below).

Cornell’s BirdNET is built on the lab’s ground-breaking work in the science of sound.

     BirdNET leads the field by a wide margin:

      The free Cornell app, built on the lab’s industry-leading bioacoustics science, is by far the most accurate. It named the right bird in every one of 10 trials. While some of the other apps are easier to use, Cornell’s BirdNET gives you a glimpse of the scientific process: it creates the digital graphic of birdsongs as you record and then lets you play a role in setting how to run the identification. Once you’ve mastered the app, the window into bioacoustics is an added benefit. (And since this review was posted, a new version of the BirdNET app has been incorporated into Cornell’s all-purpose birding app called Merlin. Here’s our review of how Merlin works with the sound ID element included.) 

     ChirpOMatic comes in a distant second:

      One of the oldest of the bird ID apps, ChirpOMatic ($3.99) identified the correct birdsong a little more than half the time. Created by the European nature developer iSpiny and then set up for North American species, ChirpOMatic is the simplest to use. It opens with a sole button that records the birdsong for 12 seconds and then delivers its verdicts.

     Two other popular bird apps were right about half the time:

Smart Bird ID, which offers a $29-a-year membership, is free if you’re willing to put up with listening to videos or doing some shopping.
Song Sleuth is an easy-to-use app that provides a list of likely birds — and was right about half the time in our tests.

     Song Sleuth, an app developed by Wildlife Acoustics with bird author and illustrator David Sibley, is the most elegant and one of the most downloaded apps. It’s filled with Sibley’s exquisite drawings and is well organized overall. Although Wildlife Acoustics is a leader in bioacoustics technology, that doesn’t seem to have helped the app with its precision.

     Smart Bird ID, which was correct about half the time as well, uses a cumbersome funding method: You have to either join a membership that costs $29.99 a year or watch a short video or visit the store before every use. As a result, the free version is frustrating to use because of those constant interruptions.

Bird Genie was wrong 90 percent of the time in tests.

     Some versions should be avoided until they’ve made progress:   

     Several other apps we tested hardly ever got the bird right. Bird Genie ($3.99), developed by Princeton University Press, was accurate about 10 percent of the time. Birdsong ID ($4.99) delivered the same weak performance. A few other apps, particularly some of the earliest versions, are too inconsistent to mention.

      What’s ahead for birdsong apps?

       The development of birdsong apps like these are still in their infancy, and the good news is that the more people use them, the better the precision will be. As users record sounds and signal the correct guesses,  their libraries of birdsongs and algorithms will expand and sharpen.

         One of the reasons that the Cornell Lab’s app is well ahead of the rest is that its staff has been working on the science the longest; they track the largest number of species and have built an entire center devoted to bioacoustics.

Birders on Washington, D.C.’s Roosevelt Island.

         While other birdsong apps are focused on the birder, Cornell has released the apps simply as a way to widen the lab’s research. When users file recordings through BirdNET, the lab is extending its reach through its birders. In the future, Cornell may use the app’s recording to study not just the bird in question, but all the birdsongs picked up through the apps. 

        “We’re only doing this for the research,’’ said Holger Klinck, who directs the lab’s Center for Conservation Bioacoustics. “We see it as a data collection and citizen science tool.’’

Holger Klinck, who heads Cornell’s bioacoustics center

      The BirdNET app complements Cornell’s eBird and Merlin birding apps that enable users to record and share their bird sightings, use photos for identification and study species.

     Serious birders will want to experiment with the best of these apps and watch how they develop. Most of the best are either free or inexpensive, and the field is continuing to develop. That said, there’s no reason all birders who like to rely on apps shouldn’t include BirdNET, eBird and Merlin in their toolbox. Not only do they improve our birding skills and provide insights into the fabric of birdsongs, they are a good way to give back to Cornell research that is so important to the future of birds.

How we tested the birdsong apps:  

We put all the apps through the same paces to test their accuracy and usability. We began by playing loud and clear birdsongs in a home setting with no other competing noise. Each app was tested with 5 common birdsongs – sometimes two or three times if the answers weren’t correct the first time – using species that could be found in our area.

The top scores were BirdNET with 100 percent correct and ChirpOMeter with 58 percent right. We calculated the results below – and then went to a playoff round between BirdNET and ChirpOMeter, using songs of less common birds. Once again, it wasn’t close. BirdNET got them all right, and ChirpOMeter identified a little more than half. We converted those into rankings on a 100-point scale on the following graph. The blue bar is the percentage of accurate guesses in our test, and the orange bar is the overall ratings the apps get from Apple users.

7 responses to “Which is the best birdsong ID app? We tested them all and have a winner.”

  1. Thanks for this review! I had not checked in on this technology for a long time because the earlier apps either failed so badly or had too few species in their libraries. I used BirdNET today (April 20, 2021) in Boise, ID, and it correctly identified 3 different Song Sparrows, an American Robin, a House Sparrow, a Black-capped Chickadee singing “May Day,” and the American goldfinch, given a group of 4 birds twittering together. It was 100% accurate.

  2. This is a wonderful review article! I hope that you’ll do a follow-up second review a year or so after your first one, to see how the field has progressed. Things to consider including:

    1) Merlin Bird ID (another free app from the Cornell Lab of Ornithology, makers of BirdNET) is getting ready to roll out a new Sound ID feature soon: https://merlin.allaboutbirds.org/sound-id/ This should be essentially a polished version of BirdNET, embedded within a larger, more general field guide app. Once the new feature is released, it would be great to hear an independent analysis of how it stacks up against BirdNET and other sound ID apps!

    2) This area of technology seems to be growing rapidly. Have any of the other apps improved? Any new apps added to the field, besides Merlin?

    3) What happens if you subject these apps to more challenging conditions? Your testing procedure in this first review only used “loud and clear birdsongs in a home setting with no other competing noise.” While that’s an excellent starting point, real-world field conditions often include quieter or most distant birds, background noise, multiple species singing at once, and so forth. Also, did you only test the most commonly-sung songs of each species, or did you also try alarm calls and other sounds, less-common song variants, regional dialects, etc.? How tough (or maybe I should say realistic) do conditions have to be before BirdNET drops significantly in accuracy — or before other apps hit 0%?

    (Full disclosure: I work at the Cornell Lab of Ornithology. I’m not directly involved in developing any of the Lab’s apps, but I’m probably still biased. It’s always great to hear an independent review!)

    1. HughAnders Gyllenhaal Avatar
      HughAnders Gyllenhaal

      Leo, Thanks so much for your note. I like your idea of an update on this, and it’s certainly warranted. I do get a very different result from BirdNET when I’m in the field, and not surprisingly it’s far less able to get the right species. Still, I couldn’t figure out how do a level test with all the apps along that line. I am interested in how the others are progressing, but haven’t had time to keep up. We’ve recently signed up to write a book on the broader issues around bird protections and conservation. We’re going to spend a week at Cornell doing interviews at the end of Augusts. We do plan to do a post on the Merlin update, and in fact just talked with Cornell folks today. What part of the lab do you work on? Perhaps we’ll get to see you in Ithaca. By the way, a better email is Andersg2021@gmail.com, in case you want to write back. Again, many thanks for your thoughtful reply.

  3. […] long ago, we wrote about the smartphone apps you can take on the birding trail if you need help identifying birds by ear. This new invention, […]

    1. Shahnawaz Ahmad Avatar
      Shahnawaz Ahmad

      Great work on your blog post! I found it to be an informative and thought-provoking read. To explore further on the topic, simply click here. Thanks for sharing your insights!

  4. Shahnawaz Ahmad Avatar
    Shahnawaz Ahmad

    Great work on your blog post! I found it to be an informative and thought-provoking read. To explore further on the topic, simply click here. Thanks for sharing your insights!

    1. Thanks so much for your note. Much appreciated.

Leave a Reply

%d bloggers like this: