Method
How Suede Cosmos is built
Suede Cosmos is an interactive music discovery map. Each of its 181 stars is an artist, and each of its 466 lines joins two artists whose MusicBrainz genres and tags overlap. This page walks through where the data comes from and how a build script turns it into clusters, connections and a fixed layout.
- Artists
- 181
- Connections
- 466
- Genre clusters
- 8
- Countries
- 17
- Cover art links
- 179
- Data built
- May 24, 2026
1. Artist data from MusicBrainz
The map starts from a hand-picked seed list of 198 artist names. For each name, the build script searches the artist names on MusicBrainz, the open music encyclopedia, and keeps the top match. It then requests that artist's record with its genres and tags, sorted by their MusicBrainz vote counts, along with the artist's country code. 181 of the 198 names were fetched into the current build. A name is skipped when its request fails, and the next rebuild requests it again. A name is also skipped when MusicBrainz lists the artist under a different canonical name, because the search does not look at aliases.
Requests run 1.1 seconds apart and each artist's result is cached on disk, so a rebuild only fetches names that are not cached yet. The finished data ships with the site as one JSON file, and the map makes no MusicBrainz calls while you browse it.
2. Cover art from the Cover Art Archive
For each artist the script lists up to five album release groups on MusicBrainz, picks the one with the earliest first-release date, and links its 250-pixel front cover from the Cover Art Archive. 179 of the 181 artists have a cover link, and the cover appears in the artist panel when you select a star.
3. Scoring connections
Every artist gets a tag set: up to 6 of its top MusicBrainz genres plus up to 6 of its top tags, lowercased. Two artists are compared with the Jaccard index, the number of labels they share divided by the number of distinct labels across both sets. A pair qualifies at 18% or more.
Each artist keeps its 4 highest qualifying matches, and a pair picked from both sides is stored once. An artist can end up with more than 4 lines when other artists pick it among their own best matches. Scores in this build run from 18% to 86%; the top score joins Travis Scott and Megan Thee Stallion, who share hip hop, trap, southern hip hop and pop rap.
The strongest connection that crosses a genre line joins Big Thief (Rock) and Iron & Wine (Folk and Country) at 83%. Their stored genre lists share indie rock, alternative rock, folk rock, lo-fi and singer-songwriter.
In the artist panel, the Nearby list shows these scores as whole numbers, strongest first. A line on the map means two artists are described with overlapping genre and tag labels on MusicBrainz. It maps how the music is tagged, which makes it a quick route from an artist you know to others tagged the same way. Connections describe data relationships, not collaborations, endorsements or influence.
4. Genre clusters and star colors
The script reads an artist's genres in order (its tags, when it has no genres) and assigns the cluster of the first one that contains a cluster keyword. Clusters are checked in the order below, so a label that fits two of them goes to the one listed first: folk rock counts as Rock, and jazz rap counts as Hip-Hop. That is why Bob Dylan, whose first MusicBrainz genre is folk rock, sits in Rock with 6 connections into Folk and Country. One artist has no genre or tag that matches and stays outside the clusters.
| Cluster | Keywords | Artists |
|---|---|---|
| Rock | rock, metal, punk, grunge, emo, prog, shoegaze | 30 |
| Hip-Hop | hip hop, rap, trap, grime, drill | 22 |
| Electronic | electronic, techno, house, edm, trance, dubstep, ambient, idm, drum n bass, breakbeat, electronica, industrial | 25 |
| Pop | pop, synthpop, dream pop, art pop, electropop, k-pop | 20 |
| Jazz | jazz, bebop, swing, fusion | 23 |
| Classical | classical, orchestral, opera, baroque, romantic, chamber, minimalism, composer | 17 |
| Folk and Country | folk, country, americana, bluegrass, singer-songwriter | 22 |
| R&B and Soul | r&b, rnb, r and b, soul, funk, gospel, neo soul, motown | 21 |
Each cluster has its own hue, and a star's brightness and size grow with the number of genres and tags the artist carries on MusicBrainz, capped at 30 and mapped onto a range from 0.3 to 1. Well-documented artists shine brighter. See all clusters side by side on the genre clusters page.
5. A layout computed once
Star positions come from a d3-force simulation run for 500 steps at build time. Each connection pulls its two artists together with a strength of 0.6 times its score and a target distance of 80, every star pushes the others away with a charge of -40, and a collision radius of 16 keeps stars from overlapping. Because most connections stay inside a genre, artists of the same cluster settle near each other.
The positions are frozen into the data file. In the browser, dragging, zooming and searching move only the camera, which keeps the canvas smooth on phones.
6. Listening previews
When you select a star, the map asks the iTunes Search API for songs under that artist's name, and the artist panel offers a play button for the first 30-second preview whose artist name matches exactly. When there is no match, the panel links to an Apple Music search for the artist instead.
Sources and licenses
Artist names, MusicBrainz identifiers and countries are MusicBrainz core data, released under CC0. Genres and tags are MusicBrainz supplementary data, released under Creative Commons Attribution-NonCommercial-ShareAlike 3.0, as set out in the MusicBrainz data license. Thanks to the MusicBrainz editors whose votes shape every tag on this map. Cover images are served by the Cover Art Archive.
The genre and tag lists, cluster assignments, connections and similarity scores shown on Suede Cosmos are derived from MusicBrainz supplementary data. They are shared under the same CC BY-NC-SA 3.0 license, with credit to MusicBrainz.