New music search engine cosine.club suggests tracks based on similarity

A new search engine, cosine.club, is giving users the ability to search for new music based on its similarity to their favourite tracks.

The AI discovery tool was launched by the creators of the popular YouTube channel hurfyd in April, and the “music similarity search engine” currently contains a database of 1,152,633 tracks.

Users of the search engine can input the track of their choice into the search bar, and then they will receive a number of recommendations based on music deemed to have similar attributes.

Read this next: Google Assistant can reportedly detect samples "less than a second long"

For example, a search of Kerri Chandler's 'Rain’ draws results: X-Coast's 'Rocky Bay’, Mallory Becer's 'Groove Me' (DJ Wasb Magic Mix)’ and Raise The Roof's 'Talking About Freedom (Dub)’ amongst others. The tracks are available to stream through YouTube and are ranked in order of similarity.

cosine.club describes the search engine as “repurposing a machine learning model originally designed to classify music genres, generating vector embeddings for over a million electronic music tracks."

Read this next: TikTok is trialling a new AI music-making tool

It does this by calculating a “cosine similarity metric" between the reference track and all the other tracks in the platform’s database.

The search engine is the latest in many AI-assisted tools for discovering new music.

Such programs have been launched by the likes of Universal Music Group, who announced their AI extension Similarity Search — as well as a platform by UK label Timehri Records called Timehri Dig Assistant, which makes algorithmic suggestions based on the input of a Discogs record.

To try out cosine.club click here

Jamaal Johnson is Mixmag's Digital Intern, follow him on Instagram