Google has updated Sound Search with some of the neural network-backed AI found in its Now Playing tool.
Now Playing debuted on the Pixel 2 in late 2017. The always-listening feature automatically recognizes songs and shows the name on the ambient display — all without pinging a server. Google was able to crunch the fingerprints of some 50,000 songs and store them on the Pixel phones without taking up much space. The result is an incredibly quick way to learn the artist and title for many popular songs. Google discovered that this song fingerprinting method was far more efficient than what it used in Sound Search, which relies on listening via the microphone, sending a sound sample back to Google’s servers, performing the search, and then returning the results to the phone.
Sound Search uses the same fingerprinting technology found in Now Playing with the full power of its server-side processing. It is accessible through the Google Search or Google Assistant apps on Android phones. Users can start a query and, if music is playing nearby, tap the “What’s this song?” tool that pops up. Alternately, people can simply ask “Hey Google, what’s this song?” Google says the new AI-powered processing on its servers is much faster and more accurate than before.
The new Sound Search should already be available to most Android phones.
