AudioWeather is a dataset used for training a generative audio model. The dataset is comprised of 44,800 audio recordings of Australian birds in thier natural enviroments. Each entry is annotated with geolocation coordinates, time of day, day of year and climate conditions.

The dataset is currrently private due to licencing negotiations, but a public facing archive of the data, filterable by all metadata can be found here: https://audioweather.com/