This new music search engine recommends tracks based on similarity

A new music search engine recommends tracks based on their similarities to others. Cosine.club’s database contains more than 1,152,600 pieces of music so far.

The “music similarity search engine” was launched in April 2024, and has been developed by the same team as the hurfyd YouTube channel.

Users can enter a track title in the search bar and receive multiple results, recommending music with similar attributes. For example, searching for Frankie Knuckles’ ‘Your Love’ will bring up recommendations for Lisa Mitchell’s ‘Rescue Me (DJ Stomp Mix)’ and No You Turn’s ‘I Still Love You (European Mix)’, while searching for Paul Van Dyk’s ‘For An Angel’ will suggest Time Motion Project’s ‘Pump This Party (Maxi Trance Mix)’.

Try it for yourself here.

Cosine.club’s functionality repurposes a machine learning model first built to classify specific genres. Vector embeds are generated for original source tracks, which are used to calculate the closest matches. The platform is similar to another AI discovery tool, Timehri Dig Assistant, which scours the Discogs vinyl data base for similar tune and was unveiled by UK label Timerhi last year. 

Last month, a new app designed to help people dig for music online also went live. When you open the Dig This app, five styles of music available on Discogs will be shown – such as Baltimore club, hip-house, electro, Italo house and dark ambient – and if you click on one of the styles, it will take you through to a randomly generated release