Google's Hum to Search: Voice is the New Text

Long are gone the days when text was the only method of searching-and-finding what you need. Today, there is a need to adjust to new methods people resort to in order to communicate; and voice is one of the fastest-growing ones on this list.

From dictating messages to Siri to asking Alexa about the weather, people increasingly treat voice assistants as integral parts of their lives -- and the industry has to keep up with the changing needs of the customers.

With that in mind, Google has introduced a new “hum to search” feature to its search -- and it quite literally lets the user hum the song they have in mind while Google's machine learning algorithm identifies the pattern they reproduce.

To make the process conversational, all users have to do is ask Google "What's the song?" and proceed with humming the tune. Google then will rank the search results based on how close of a match they are to the music, giving the user a few options to look into to find the best and closest match.

According to Verge, the feature uses Google's machine learning models to transform the audio into a binary sequence of numbers that encode the song’s tune, which makes it easy to compare against existing songs on the web. The algorithms have been trained on such sources as real-life whistling or humming of people trying to sing a song, as well as studio recordings by eliminating the sound of instruments and vocal enhancement in order to isolate the singer's audio.