A reasonably simple script that generates a list of bird calls detected with birdnetlib, nicely formatted for use with Audacity.
For now, please install from GitHub (only tested on Linux, but in theory might work elsewhere).
pip install git+https://github.com/kdm9/birdnet-labeller.git#egg=birdnet-labeller
Ideally, do this within a virtual environment, or with pipx.
$ birdnet-labeller --help
usage: birdnet-labeller [-h] [--output OUTPUT] [--longitude LONGITUDE] [--latitude LATITUDE] [--date DATE] [--min-confidence MIN_CONFIDENCE] input
positional arguments:
input
options:
-h, --help show this help message and exit
--output OUTPUT, -o OUTPUT
Output labels.txt file for Audacity
--longitude LONGITUDE, -x LONGITUDE, --lon LONGITUDE
Observation longitude (default: Tübingen-ish)
--latitude LATITUDE, -y LATITUDE, --lat LATITUDE
Observation latitude (default: Tübingen-ish)
--date DATE, -d DATE Observation date/time
--min-confidence MIN_CONFIDENCE, -c MIN_CONFIDENCE
Minimum model confidence to report
For example, from recent recording: birdnet-labeller -o 240816_001.labels.txt --date 2024-08-16 -x 9 -y 48 -c 0.3 240816_001.WAV, which generates
a .labels.txt file corresponding to a WAV straight from a field recording.
Then, in Audacity: File -> Import -> Audio -> Select the wav file, and File ->
Import -> Labels -> Select the labels.txt file. You should see the labels pop
up at the correct times.
There are a bunch of GUIs for Birdnet analysis, however I typically want to edit and compose soundscapes in Audacity rather than just find a list of birds. So, I needed a way to import birdnet calls into audacity, hence this script. Other, possibly better, solutions probably exist; this was largely a learning exercise that ended up being useful.