search capabilities in Google They are practically unlimited. We are used to typing our queries in the search bar, but new times have introduced the voice assistant, which will attend to our requests and show us the most relevant results. This even applies to songs, even if you can only hum them.
Has it ever happened to you that you have a catchy tune that won’t get out of your head and you want to know the title or the artist (surely so that later it doesn’t stop playing in a loop on your Spotify)?
If we had a fragment of said musical theme, we could always use one of the many apps that there are to find songs, with Shazam at the forefront. If not, we can always put our shame aside and start whistle or hum the song (or sing it as well as you know) for Google.
The trick to search song
Google’s voice assistant is capable of find any song in the world just by humming it for 15 seconds. You just have to press the microphone button and start singing. The process can be done by the Google app or the search widget.
By default, the search is for words, but if we look at the bottom margin, it will appear «search song“That’s our section. After a few seconds listening to you whistle or hum, Google will proceed to search with its algorithm for matches with your hums, and will show you a table of results.
You can select the best match and browse information about the song and artist, view the accompanying music videos, listen to the song on your favorite music app, find the lyrics, and even view other recordings of the song when they become available.
In Google Assistant, it’s just as simple. Gave “Hey Google, what’s this song?” and then hum or whistle the tune. The next time you can’t remember the name of that catchy song you heard on the radio or that classic your parents love, just start humming. you will have your response in record time.
How does the algorithm work?
An easy way to explain it is that the melody of a song is like your fingerprint: each has its own unique identity. Google has created machine learning models that can match your humming, whistling, or singing to the correct “fingerprint.”
Those machine learning models transform the audio into a sequence based on numbers which represents the melody of the song and are trained to identify songs based on a variety of sources, including humans singing, whistling, or humming, as well as studio recordings. The algorithms also remove all other details, such as accompanying instruments and the timbre and pitch of the voice.
What remains is a sequence based on song numbers, or that fingerprint. Then, those sequences are compared to thousands of songs from around the world and we identify possible matches in real time.