TECH
YouTube’s new experiment lets users hum to find songs: A game-changer in music recognition
By humming, singing, or recording at least three seconds of a song, users trigger the platform’s machine-learning algorithms. The algorithms then match the audio input to the song’s “fingerprint” or signature melody, directing users to related YouTube content, which might include official music videos, user-generated clips, or Shorts.
New Delhi: YouTube has unveiled an innovative experiment on its Android app that could revolutionize the way we identify songs. Unlike traditional methods, YouTube’s new feature allows users to hum, sing, or record a song to search for it within the app. This leap forward in music recognition could outshine competitors like Shazam and even Google’s previous efforts.
This song search feature is currently being tested with a select group of Android users. Those fortunate enough to access it can switch from the regular voice search to this new song search option. By humming, singing, or recording at least three seconds of a song, users trigger the platform’s machine-learning algorithms. The algorithms then match the audio input to the song’s “fingerprint” or signature melody, directing users to related YouTube content, which might include official music videos, user-generated clips, or Shorts.
YouTube’s experiment reflects a shift towards more intuitive song discovery. This initiative, if expanded to a wider audience, could significantly simplify the process of identifying music, given that YouTube is a popular destination for song searches.
Interestingly, this concept isn’t entirely new for Google. Back in 2020, YouTube’s parent company, Google, introduced a similar capability across its app, Google Search widget, and Google Assistant. However, Google’s feature required users to hum or sing for a longer span—around 10 to 15 seconds—before identifying a song.
The technology underpinning this new feature is rooted in machine learning models. These models analyze the audio input’s unique characteristics and match them with a database of known song “fingerprints.” The technology is similar to what powers Google’s original hum-to-search feature, a YouTube spokesperson confirmed.
While music recognition tools like SoundHound and MusixMatch also offer hum-to-search capabilities, YouTube’s colossal user base and Google’s technology integration could give it a significant edge in this field. As music enthusiasts eagerly await this new feature’s broader release, it’s worth keeping an eye on how it reshapes the way we interact with music discovery in the digital age.