Chirpity: Identify Birds by Sound

Using Machine Learning to identify bird calls in audio recordings

Chirpity screenshot showing spectrogram, results table and context menu

Welcome to Chirpity, your ultimate companion for identifying bird vocalisations. Designed with Nocmig enthusiasts and bioacoustic researchers in mind, Chirpity revolutionises the process of reviewing extensive audio files for the presence of avian sounds. Powered by cutting-edge Machine Learning, Chirpity offers the choice between the well-known BirdNET and a bespoke Nocmig model - finely tuned for Nocturnal Flight Calls - ensuring identification tailored to your specific needs. With intuitive features for reviewing, tagging, and organizing detected calls, Chirpity saves valuable time and effort. Available for Windows, Mac and Linux platforms, download Chirpity today!

“ Chirpity is truly a game changer for me. Now I'm not afraid of recordings for entire weeks because I know I can easily and quickly screen through what I'm interested in! ”

Stanislas Wroza

Key Features

Platform and Languages

Versions are available for Windows, MacOS (both M-series and Intel) and Linux.
The interface can be translated into any one of 11 languages
Birds' common names are available in 28 languages.

Save Results

In-built Archive: Save detections to an archive database for future reference and retrieval.
Export clips: Extract audio highlights, and export them in your chosen format and quality.
Export Records: Export detections as Audacity labels, a Raven selection table, in eBird Record format or as a CSV.
Locations: Set custom recording locations, filter saved records by species, location, date and time.
Charts: View charts of saved records. Visualise the change in numbers of species' detections through the season.

Media Capabilities

Broad file format support: Supports WAV files in 32bit float PCM and those larger than 2GB (RF64). You can also analyse and save common compressed file formats.
Support for BEXT and GUANO metadata: If GUANO metadata is embedded in the file, this is read, and the recording location and timestamp are set automatically
Enhance audio: Audio filters allow you to adjust background noise.
Organise files: Organise your source recordings into a structured audio library.

Analysis

Compare detections with reference recordings from Xeno-Canto — Compare detections with reference audio files.

Compare results with reference calls: Compare any species detection with reference calls sourced from xeno-canto , sorted by call type.
Background Analysis: Analyse audio in the background while you continue to explore the timeline.
Retrospective adjustment: Adjust the detection list or confidence threshold either before or after analysis.
Easily Refine Results: Update the details of records, amend or delete them. Add records manually.
Enhance Sensitivity: Highlight and analyse specific segments for more detailed results.
Custom lists: Limit reported detections to birds in your location at the time of year, birds active at night, or provide a custom list to specify precisely the species of interest to you.
Time of day and daylight detection: Detections are tagged with the exact time of day and dawn / dusk boundaries are calculated automatically.

Performance Comparison

Many people familiar with BirdNET have used the "BirdNET Analyzer GUI" for automated bird call identification. Due to many optimisations made in the application, Chirpity is able to perform the analysis significantly faster. The following benchmark data show the time taken to analyse a file just over 3 hours long. Firstly using the native Chirpity detection model, then BirdNET within Chirpity, and finally BirdNET using the BirdNET Analyzer. The number in brackets shows how much faster than real-time this represents:

**Machine Specs:** *1.CPU: 8th Gen, Core i5 8400, 6-core "Coffee Lake", GPU: Nvidia 1660 Super. 2. 2022 M2 Macbook Pro 13", 8GB RAM*
Model	App Launch	Core i5 CPU	Core i5 GPU	M2 CPU	M2 WebGPU
Chirpity native	< 3s	80s (140x)	28s (388x)	67s (167x)	44s (252x)
Chirpity BirdNET	< 3s	106s (106x)	270s (41x)	75s (148x)	N/A
BirdNET Analyzer	> 100s	549s (20x)	N/A	126s (88x)	N/A