Chirpity is an easy-to-use, accessible app developed to aid the successful identification of bird calls within acoustic recordings. It is the ideal software solution for anyone interested in identifying bird vocalizations, whether you’re a dedicated Nocmig enthusiast or a professional bioacoustic researcher, and streamlines the often tedious process of reviewing audio recordings to quickly and accurately detect avian sounds. It features an array of intuitive features to assist with identification, including call review, tagging and background analysis.
We recently had the opportunity to speak to founder Matthew Kirkland about the software, including how he became interested in both nature and apps, why he developed Chirpity, its uses, capabilities and more.
Firstly, could you tell us a little about yourself and how you became interested in both the natural world, and app development?
I got the nature bug as a kid from my grandparents. I would spend the summer holidays in Northumberland with them and they were both outdoorsy types. We’d visit nature reserves where my grandad would point out the birds, and my nan would know the flowers – and often have a story about their names and medicinal uses, plants like speedwell and eyebright. Nonetheless, it was the birds that really ignited my passion. I think I thought I was better at bird ID than I really was, because looking back at my childhood field notes, I recorded some pretty outlandish sightings!
Now I’m a little bit more grown up andI would say I’ve become quite adept at identifying birds by sight, or ‘jizz’ (general impression, size and shape). However, one area that still befuddles me is bird song. My grandad was deaf – he piloted catalinas (flying boats) in WWII and the engine noise affected his hearing ever after. He couldn’t put me on to bird song or calls as a way to find the species around you, and I never learned the calls. Not deterred by this, after reading an account of a local birder’s Nocmig records (calls made by flying nocturnal migrants) on our local WhatsApp group, I decided to plug a cheap mic into my iPhone, and leave it recording out of my bedroom window overnight one April evening – just to see what it might pick up.
The next day, I downloaded the file onto my PC and opened it up in the sound software Audacity. This allows you to view a spectrogram of the audio, and with this view, you can scroll through the night’s recording to see the sounds. Bird calls show up as specific shapes in this view: they have a signature squiggle. After perhaps an hour of scrolling I came across something that looked like this:
What was it? I had no idea! So I sent a clip of the sound to the WhatsApp group. The county Bird recorder came back with the message ‘It’s a Whimbrel’. Well, that had me hooked! I live in suburban Luton, 50 miles from the sea at least, so having a Whimbrel over the house was astonishing.
What followed over the next weeks and months was a series of ‘What’s this?’ submissions to the newly created county Nocmig group on WhatsApp.
All the while, I was thinking there had to be a better way to sift through hours of audio. I had a computer background of sorts: I’d worked for BBC Online as a web developer many years ago. I had also completed a Masters’ degree in cognitive science back at uni and retained an interest in artificial intelligence ever since. In part because of this, I’d heard of BirdNET – which is AI software trained to recognise bird vocalisations. It comes from the same group at Cornell University that do the Merlin App. I got this to work on my audio files with reasonable results: I could open a terminal on my computer and type a command for BirdNET that would scan the file and spit out a list of bird detections at specified times in the audio. I could then go back to Audacity and see whether the predictions were correct. When they were, it was amazing, but all too often they weren’t.
For those unfamiliar with Chirpity, can you give an overview of its purpose, uses and capabilities, and explain what you were hoping to achieve in developing this application?
Well, to be honest, my initial experience with BirdNET for Nocmig was somewhat tedious. Chirpity is the result of an effort to make Nocmig simple and enjoyable – even for people who couldn’t tell if they heard a Moorhen or a motorbike!
The application lets you drop your audio recordings into it, you hit ‘Analyse’ and it scans for birds. Whenever it finds a call or song, you’ll see the result appear in a table and you can click on it, play the clip or edit the record.
Since AI Bird ID isn’t perfect by any means, there are dubious identifications. Chirpity manages this in three ways:
- You can restrict the reported species to those likely at your location, either by using an automatically generated ‘Local Birds’ list or a ‘Custom’ list with only the species you have selected.
- You can compare the predicted species with examples that have been uploaded and vetted by experts on the Xeno-Canto website.
- It’s also really easy and quick to delete, correct or annotate results before you save them to your personal archive – a database of records you can revisit, view and replay.
There is an option to view charts of the species you’ve recorded, so you can see when the peak passage is in the year and when to be on the lookout for Redwings, Common Scoters, Tree Pipits –species which are regular night migrants over most of the country. Here’s an example showing my Redwing records in 2022:
I’ve been lucky enough to pick up Stone-curlew too on several occasions, with this astonishing record being the best of them:
As you can see, with Chirpity you can adapt the visualisation and export clips from your recordings.
However, Chirpity isn’t just for Nocmig – many of its users are recording in the field with remote field recorders, like Song Meters you sell. With external power, some of these recorders can be rigged to run 24/7 for periods of up to a month before their audio is retrieved from the SD card. These researchers use Chirpity to review the audio, check the results and export their findings either as a spreadsheet or in an eBird Record format.
It really stood out to me that Chirpity is incredibly accessible in comparison to its desktop counterparts. Was accessibility a key driver for you?
Absolutely! There are two key thoughts I have in mind when thinking about Chirpity’s development:
- What’s the best way to make this feature easy to understand and use?
- Is there a way it can be delightful?
All the settings have a help icon next to them so you can read what they do, and there are lots of keyboard shortcuts to speed things up once you get used to the application. And for non-English speakers, the interface can be displayed in any one of 11 different languages!
What challenges did you face when developing this application, and how did you overcome these?
Well, one thing I have learned is that software isn’t soft, it’s hard! There are so many challenges I’ve had to overcome, it’s difficult to know where to start. For its users, the one I am most pleased to have resolved is getting an insight into how people actually use the application, where they struggle and what doesn’t work. Most of Chirpity’s users aren’t technical people, they won’t report a bug – they just think they messed up. I now have a system that shows me, anonymously, what people do with the application, if it has errors and if so, what someone did to cause it. That means I can fix it! Another challenge has been managing updates. The application is available for Windows, Mac and Linux computers, so I was very pleased to find a solution that packages each version, tests it and alerts users to the update automatically. Finally, training a call recognition model that has good results in real world use – like the Nocmig model included with Chirpity – is really hard. It took me about a year to make it effective – I had to learn the AI platform from scratch but even so, most of the time was spent preparing the 500,000 or so examples in the training data!
Given your active presence in both bird and software forums and your engagement with users, could you share which Chirpity project has been the most personally rewarding for you?
Well, I’m really quite proud of the fact that the County bird recorder that identified that Whimbrel at the start of this journey now uses Chirpity for his own Nocmig analysis! I’m also amazed that it’s taken off so widely on the international stage. I have a map of Chirpity user’s locations, it amazes me that it’s got this much traction when it’s been less than a year since I put the first version out to the public (and I don’t do any paid advertising):
Technology is constantly being updated and improved – what future developments can we expect for Chirpity?
There isn’t what you might call a roadmap for new features – things move and change too quickly. I will keep to the vision – simple and delightful to use. The folks at Cornell (the BirdNET people) are teasing a big update for this year, and we’re in touch over ensuring that update will work with Chirpity when it lands.
I’d like to add a self-learning element to the application, so it remembers and adapts to corrections – though this is a challenge that is yet to be overcome!
I’ve recently added a “Buy Me a Coffee” link for donations. With sufficient funding, I’ll be able to pay for a digital signature, which will clear the hurdles when installing unsigned software and allow automatic updates for Mac users. That funding will also help me cover the cost of training better Nocmig models and well, drink more coffee, right?
Are you working on any other projects that we can hear about?
Honestly – this is plenty to be getting on with! Although, I’d quite like a version of Chirpity that works like ChatGPT – where you give it a file and can say “What birds are in this? How do you know? Let me listen to the Stone-curlew you found…That’s great! Send it to my socials”
Find out more and download Chirpity today here.
