Spotify Universe

Spotify Universe

Links:

Site

Summary

Spotify Universe is a 3d visualisation of relationships and similarities between ~1.4 million artists and ~6000 genres on the Spotify music streaming platform. The project is currently in development and is being built using React, React Three Fiber, WebGL, Python, MariaDB, MongoDB and SQLite.

Check it out here: https://universe.davitg.com

The project has gone through multiple phases of rewrites and toolchain modifications, but is now on it’s way to becoming a MVP sometime in the near future.

Difficulties

This is by far the hardest and most complicated project I have ever attempted. Each little detail of the project is expansive and needs creative solutions to solve problems. A few examples of problems I have had to face in this project includes the handling and efficient storage of data from Spotify API, writing an efficient browser-based instanced text rendering solution using image atlases and MSDFs, giving labels to partially labelled data (labels being genres in this case), and artist / genre colouring based on dimensionality reduction algorithms.

Similar Projects

Not many projects exist like this one. The popular Every Noise project is the most similar, written by Glenn McDonald (one of the machine learning engineers currently working at Spotify). Spotify Universe was heavily inspired by his work, but my project tries to reinterpret the data and make interaction and music discovery easier, more intuitive and more informative. The original genre labels that Spotify comes up with for their tracks is chosen by algorithms directly written by Glenn McDonald, so I just want to say I hugely respect their work.

@ Davit Gogiberidze 2023