Google’s search tools are getting a new feature called ‘hum to search’, which is going to use machine learning to identify a song. A user can simply hum the song into their device’s microphone and ‘hum to search’ will show them the song results.
Both iOS and Android users can access the feature in the Google app. On Android, the feature works in 20 different languages, while it only works in English on iOS for now. ‘Hum to search’ is also available on Google Assistant.
The feature also works if a user asks Google ‘'What’s the song?’', or they can choose to hum the song after tapping the ‘search a song’ button.
The search results will show the closest possible matches of what the AI identifies the song to be. The user can then tap the results to listen to the song.
As explained by Google, the new feature functions on machine learning models that “transform the audio into a number-based sequence representing the song’s melody,” which is then compared against existing songs.
According to Google, the machine learning models use a variety of sources such as humans singing, whistling, humming, and studio recordings as well. The instruments or vocal quality of these audio sources are then stripped away to focus on the numeric sequence.
The feature works regardless of a user being tone deaf or having perfect pitch.