Finn's Library

About this map

Albums from my beets library, placed by how they sound. An audio model (Discogs-EffNet) analyzed each album and converted their sound into high level vectors. Similar-sounding albums sit close together, and lines connect the closest matches.

Select an album to see the calculated numbers. Try the natural language tool to describe a sound to highlight matches. It's neat. Hit the gather button :-)

Phrase search runs locally on my desktop, turning your query into the vectors. Matching against the library happens in your browser.

Map appearance

Album artwork 100% Size albums by Color albums by

Describe a sound

Strong matches grow and brighten; other albums fade. Your filters still hide excluded albums.

Your phrase goes to this server to be turned into numbers; the matching happens in your browser.

Sound lens and filters

Sound lens and filters Find a label Danceable: minimum 0.00 Danceable: maximum 1.00 Vocals Include unknown filter values

Danceable is the Essentia classifier probability. Non-matches are hidden and frozen; visible albums keep moving.

Show group labels

Connections

Visible connections

Hover or select an album to reveal its connections. Its closest neighbors stand out; quieter bridges connect sounds. Click a group label to explore its albums.

Limit nearest neighbors Neighbors 8 Filter by similarity Minimum similarity 0.70

When both are on, similarity filters the nearest neighbors. Connections work in both directions.

Motion & rendering

Layout

Show albums as Auto: minimum zoom 1.00

Small covers stay cached; visible albums gain detail as you zoom in.

Loading overview covers.

Color key

The frame around each cover shows its group.

Your library

On the public site this is the host's published library. Open your own export to replace it — files you open stay in this browser tab.

Open album export

No library loaded

Open an album export to explore its connections.