Google Introduces Hum to Search to Identify Stuck Songs
Google now allows users to hum, whistle, or sing into the Google app or Assistant to identify mysterious songs. Machine learning algorithms strip away pitch and tone to match the melody against a massive database.
Google introduces a new "Hum to Search" feature that allows users to identify songs stuck in their heads. People simply tap the microphone in the Google app for Android or iOS and ask "what's this song?" or tap the "Search a song" button to start.
After a user hums, whistles, or sings for 10 to 15 seconds, machine learning models transform the audio into a number-based sequence. These algorithms remove details like pitch, accompanying instruments, and vocal timbre to isolate the exact melody fingerprint.
The system then compares this sequence to thousands of songs in real time and displays potential matches with cover art, artist names, and match percentages. This capability builds on previous audio recognition technology and currently supports over 20 languages on Android and English on iOS.