"mobile-device"

The models are trained to identify songs based on a variety of sources, including humans singing, whistling or humming, as well as studio recordings.