Google launches new ‘hum to search’ feature to help you identify that song stuck in your head
Tech

Google launches new ‘hum to search’ feature to help you identify that song stuck in your head

Users can tap the mic icon on the Google app and ask, “What's this song?” and hum it for the software to suggest potential song matches. This is also available on Google Assistant.

   
Google headquarters in California | Wikimedia Commons

Google headquarters in California | Wikimedia Commons

New Delhi: Google Inc. Friday added a new feature that allows users to search for a song by humming or whistling the tune for 10-15 seconds.

The new feature can be accessed in different ways. Users must tap the mic icon on the latest version of the Google app or search widget and ask, “What’s this song?” and hum it, or click the “Search a song” button. If using Google Assistant, one simply asks, “Hey Google, what’s this song?”

After they hum the tune, which need not even be pitch perfect, Google’s machine learning algorithm finds potential song matches. The feature is currently available in English on iOS and in more than 20 languages on Android, though the company “hope[s] to expand this to more languages in the future.”

Google also uploaded a promotional ad on its YouTube channel Friday, which showed clips of various people humming and whistling the lyrics of pop song Dance Monkey by Australian singer-songwriter Tones and I.

The company also made the announcement on its official Twitter handle Friday with a common question people have when there’s a song stuck in their head: “You guys — what’s that song that goes like la la lalalala laaaa lalalala lala lala lala”.

There are other apps that already offer the hum-recognition feature like SoundHound, Shazam and TuneFind.


Also read: Google plans to turn YouTube into major e-commerce hub like Amazon, Alibaba


AI for music recognition

On its blog, Google explains how a song’s melody is like a “fingerprint” and their machine learning model can match a tune to the right fingerprint. To form this fingerprint, the audio is converted into a number-based sequence that represents the song’s melody.

The fingerprint is then compared with tons of digitised songs to find a few that appear similar and the machine learning model has the ability to match a studio-recorded version of the song, to a person’s hummed audio.

The new feature is part of Google’s long-standing project of using artificial intelligence for music recognition. The search giant launched Now Playing on its Google Pixel 2 phones in 2017, “using deep neural networks to bring low-power recognition of music to mobile devices”. A year later, it used the same technology for its SoundSearch feature in the Google app.


Also read: PlayStation inventor starts new career making robots for no pay