A music knowledge explorer
A half remembered song.
Distant drums draw you deeper.
Orpheus brought his lyre. Bring your curiosity.
36,000 artists · 2,700 styles
Enter
Search
A music knowledge explorer
A half remembered song.
Distant drums draw you deeper.
Orpheus brought his lyre. Bring your curiosity.
36,000 artists · 2,700 styles
Enter
Search
Relics
According to legend, Orpheus was the greatest musician of his time. He was forced to rescue his beloved Eurydice from the underworld. This maze represents his experiences as he explores the underworld, perfecting his craft in preparation for his ultimate musical challenge.
Enter Orpheus' musical maze:
Each room is a single musical style ("Progressive Bluegrass" or "Math Rock") or artist ("Wolfgang Amadeus Mozart" or "David Bowie"), reconciled from many data sources. Collectively, the tens of thousands of rooms form a maze to explore.
Search bar (also top right) — not modal, so you can type the name of either a style or an artist. Has fuzzy match and autocomplete. Hitting enter selects the top choice.
Room exit buttons (in organized panes on left, as links in text on right) — Math Rock red buttons take you to a style room; Bowie gold-and-country-colored buttons take you to artist rooms. Buttons for influenced / influenced by, members, member or past member of, and spinoffs are placed below the text area when available.
A relic is a room you've saved. Saved relics show on the landing page in the Relics pill row. From the Relics page (click the row), you can weave them into golden threads — named, ordered, shareable subsets you curate by hand.
Weaving threads — on the Relics page, The Skein is your default catalog: every relic not yet in a named thread. Below it, your golden threads. Drag relics between strips to curate — every relic lives in exactly one strip at a time. The bright lyre stamp (left of The Skein) creates a new thread. Each named thread shows a smaller lyre stamp just left of its name — click to rename — and a ring stamp on the right to delete.
Finding a relic — the Relics page header has its own search (limited to your saved rooms) and a random button (picks from your own collection). Useful when the Skein has grown.
Sharing a thread — every named thread has a Share button that generates a URL encoding its name and contents. Threads are anonymous; recipients see only what's inside, and can adopt the thread into their own Relics with one click.
I'm Siegfried Martens — a data scientist who loves music and elegant visualization. As an explorer, I've always wanted a tool that gave me the bigger picture: how artists and styles connect, where to jump in at random. I also wanted to see what I could build from public data sources. With Orpheus, my playlists are always on the move.
Orpheus is phase 2 of the Daedalus projects. Phase 1 is Ariadne, a Tableau dashboard view of the same data: stylish_networks. For more about me, see s-martens.com.
I built custom crawlers that spent a long time exploring Wikipedia, Wikidata, MusicBrainz and Spotify's public APIs, and after a lot more entity resolution, curation, and knowledge graph building, I now have a data set that organizes a clean interface to these disparate sources. The data is not publicly available in download form, though I am exploring releasing cleaned subsets.
Not yet. Orpheus is currently a static site — every room is pre-generated as a JSON file. The next-generation backend, called Lyre, will bring a proper query API and dynamic data. Until then, what you see in the browser is what's available.
Both rooms give you consistent metadata in the sub-head:
Both also provide organized exit buttons on the left (styles at top, artists below), a text pane with content from the corresponding Wikipedia page (with embedded exit links), and album art.
The sub-head adds information about when the style originated and where; tooltips tell you where the info came from. The exit pane on the left has, at the top, links to the style's origins in earlier styles and what styles it influenced. At the bottom, the artists associated with the style are organized by time period — precursors, founders, early years, middle period, later years, recent, and date unknown. Not every category appears for every style. Artists are colored by country of origin (top 7 countries; the rest are grey for "Other," with the country always shown in the tooltip). Within each time period, artists are grouped by country, with a flag on the first artist in each country group, and sorted by score within the group.
Style rooms also show up to four randomly selected album covers from associated artists. Clicking a cover exits to the corresponding artist room.
The sub-head adds When (years active), Where (country of origin), and either Born + Died (for people) or Formed (for bands), plus Occupations, Aliases, and Instance of when known. The exit pane on the left has Artist Styles at top and Related Artists below — grouped by the styles they share with this artist, then organized within each group by country. The same per-list shuffle button (small maze tile) sits inline with the Related Artists heading. The color legend at the bottom gives summary stats by country. The artist room also features the discography, inset at right: the art for the selected album sits above the full discography, navigable via the entries or the left/right arrows on the art. Clicking the art enters a zoomed mode at the highest resolution available.
Two artists are "related" if they share styles — but with a twist. The match uses Tanimoto similarity (a kind of overlap measure), weighted by TF-IDF: artists who share a rare style ("Norwegian Black Metal") count for a lot more than artists who share a common one ("Rock"). So Mayhem and Darkthrone end up close even though both also touch giant generic buckets that everyone else lives in. It's a surface match — it doesn't know who actually sounds like whom — but it tracks scene and lineage well, which is usually what you want.
Two reasons. First, Wikipedia and Wikidata's style taxonomy is mostly organized around popular music — "Classical" itself is treated as a single bucket, while "Indie Rock" gets sliced thirty ways. Periods and forms (Baroque, Romantic, opera, symphony) don't behave like genre tags, so the maze flattens them. Second, the classical greats have hundreds of recordings of the same works — modern reissues, box sets, conductor variants — which inflates discographies and skews album-cover sampling. Spotify popularity (one of the score inputs) also undervalues classical relative to its cultural weight. Fixing this properly is on the roadmap; for now, treat the classical corner of the maze as gestural.
At the moment the data crawlers are manually driven. Automating the ingestion step is one of the next things on the list.
This app features language from Wikipedia. More importantly it features album covers of many types, just like in a record store — some titles or cover art are not appropriate for children. Browse accordingly.
The data is not authoritative. It reflects what Wikipedia, Wikidata, MusicBrainz, and Spotify happen to know — which means scenes with strong online communities are well-covered, and others less so. If your favorite artist isn't here or sits in the wrong room, that's the data, not a judgment.
Orpheus stands on the work of:
Thanks to the volunteer editors and contributors who build these resources. Orpheus is a derivative work; nothing here would exist without them.
Built by Siegfried Martens — s-martens.com.