Loading listeners…
Everything above shows your tracks placed in 2D using precomputed embeddings. Pick a listener to animate their plays, recolor points by any song attribute, and click points for details as you go.
Each song in your dataset has many attributes (danceability, energy, tempo, etc.). Projections such as PCA, UMAP, or t-SNE compress all those attributes into just two axes so we can see the entire catalog at once. PCA stands for Principal Component Analysis, t-SNE is t-distributed Stochastic Neighbor Embedding, and UMAP is Uniform Manifold Approximation and Projection. Each prioritizes different structure, letting you compare how the same songs organize under multiple views.
Want a deeper dive? This short presentation walks through the projections step by step, including how the embeddings were produced and what to look for when interpreting them.
Each dot is a track. Colors come from the attribute you choose on the right, and the path shows the selected listener's session over time.
See how this listener's songs cluster in the embedding — essentially their taste profile across the selected projection.
