How Streaming Services Automatically Build Custom Radio Stations

A method for streaming services to create custom music radio stations by analyzing genre percentages and artist relationships to pick the best songs.

What it covers

This patent describes a system that builds a custom music station based on a 'seed,' such as a specific song or artist. It creates a 'station descriptor profile' that acts as a blueprint, breaking down the desired music into specific genre percentages. The system then compares candidate songs against this blueprint using a similarity score, while also applying 'boost values' for factors like language, artist relationships, or release dates. Finally, it selects the songs with the highest combined relevancy scores to populate the station library.

What it doesn't cover

• —Does not cover manual playlist creation where a human selects every song.
• —Does not cover simple random shuffling of a user's entire library.
• —Does not cover hardware-based audio processing or signal compression techniques.
• —Does not cover social media-based music sharing or peer-to-peer recommendation systems.

The clever bit

The system uses 'focus genre profiles' that assign specific percentage weights to genres, allowing the algorithm to maintain a precise stylistic balance in a station rather than just picking songs that sound vaguely similar.

Why it matters

This technology is fundamental to the 'radio' features found in modern music streaming services. By automating the creation of genre-balanced stations, services like Spotify or Pandora can keep users engaged without requiring them to curate their own music, which is essential for the subscription-based streaming business model.

Real-world examples

• 1.Spotify Radio
• 2.Pandora stations
• 3.Apple Music station generation
• 4.YouTube Music radio features