How ryser.id identifies tracks
By Ryan Poser, who builds ryser.idUpdated First published
The short answer
ryser.id listens to the audio: it cuts the set at its transitions, fingerprints every part several times and checks each match against the set's own tempo and key. It takes no names from 1001Tracklists or from earlier analyses of the same set. Every row gets a mark that says how sure it is, and what the audio cannot name stays unknown instead of guessed.
In short
- Works on
- YouTube, SoundCloud and Mixcloud links, or an audio file up to 500 MB, 1 GB on Opener, 1.5 GB on Headliner and 2.5 GB on Studio.
- How
- The set is cut at its transitions and every part is fingerprinted several times, edges first; tempo, noise and preview checks handle blended and pitched parts. Every plan gets this, free included; Headliner and Studio also ask a second recognizer about the parts the first can't name.
- Names from
- The audio only. No names from 1001Tracklists or from earlier analyses of the same set, so what the audio cannot name stays unknown.
- How sure
- Every row is Proven, Found, Weak, To check or Unknown. Unknown rows are never filled with a guess.
- Accuracy
- Best on released music: a classic house or trance set comes back nearly complete. Edits, bootlegs and unreleased IDs are often marked unknown.
- Speed
- A one-hour set takes about 5 minutes from paste to done, a two-hour set about 10, and tracks appear while it runs.
From link to tracklist
- Fetch and cut. The set is fetched from the link you pasted (or taken from your upload), turned into audio and cut at its transitions into parts, each heard in windows of about ten seconds. Audio from a link is deleted right after the analysis; your own uploads are kept 7 days, or for as long as you pay for a plan.
- Recognise. Every part is fingerprinted several times. On a long part the first and the last window go first: when they agree, the part is one record, and when they differ, the windows in between find where one record hands over to the next. What stays silent is tried again with tempo and noise passes, and a doubtful match can be checked against a preview of the record. This is the same for every set, free or paid. On Headliner and Studio, the parts still without a name then go to a second recognizer, and a name it proposes is kept only once it checks out against the record; otherwise the part stays unknown.
- Combine and check. Answers from neighbouring windows become one track with a start and an end. Windows that agree make a match stronger; a single window alone makes it weaker. BPM and key are measured from the set and compared with what the match claims. A record at the wrong tempo, in the wrong key, or released after the set was played is set aside, never silently kept.
- Mark. Every row gets a mark that says how sure ryser.id is (below). What the audio could not name stays unknown, at its own time in the set: no name is taken from 1001Tracklists or from an earlier analysis of the same set to fill it.
- Publish. A set you make public, and a library set that names enough of itself, gets a page with the tracklist, a page per track and artist, and where to buy each record.
What the marks mean
| Mark | Meaning |
|---|---|
| Proven | The record itself was lined up against the mix, and it fits. |
| Found | More than one part of the mix agrees on this name. |
| Weak | Only one part of the mix, a doubted version, or another source names something else here. |
| To check | Heard, but not sure enough to count. A listen and a Correct confirm it. |
| Unknown | No name yet. In your own sets these rows are watched: you get a mail when one gets a name. |
On a computer, hover a row's mark for the reasoning behind it: how many parts agreed, or what was heard and why it was set aside. Reading the setlist covers the rest of a row.
Library sets
The set library picks the most watched new DJ sets from public channels every day and identifies them automatically, the same way as any other set. Nobody checks those tracklists, so a library page says it was automatically scanned and is unchecked, and shows how much of the set was identified. A library set only gets a page once its named tracks cover at least half of the set's time, with at least 8 tracks. Anyone signed in can suggest a name for a row with Fix this ID; suggestions on library sets are reviewed every week. Library sets from festivals also appear on the festival pages.
What it cannot do
Fingerprinting names released music that is in a catalogue. It has blind spots, and so does every other tool:
- Unreleased music: promos, dubplates and the DJ's own new tracks are in no database. They stay unknown, which in a tracklist reads ID - ID. See what ID means.
- Heavy blends and pitch: two records playing for a minute at once, or a record played far from its tempo.
- Edits and versions: the song can be certain while the remix or edit is not. That row is marked Weak until somebody checks it.
ryser.id publishes no accuracy percentage. Every row says how sure it is instead, so you can see for yourself which ones to check.
Corrections
The owner of a set can correct every row, and every change can be undone. On a public set, anyone signed in can confirm a row or suggest another name with Fix this ID; the owner decides. Crowd IDs explains how names from other DJs are weighed. A correction teaches the next set that plays the same record, in anyone's account.
Related
Questions
Does ryser.id copy tracklists from 1001Tracklists?
No. ryser.id listens to the audio and names tracks from that alone. It takes no names from 1001Tracklists or from earlier analyses of the same set. Older rows that came from a 1001Tracklists tracklist stay labelled as such.
Is recognition better on a paid plan?
Every set gets the same first recognition and the same check against the record, free included. Headliner and Studio also ask a second recognizer about the parts the first can't name, and keep a new name only once it checks out against the record. Paying also buys hours, a place ahead of free sets in the queue, uploads kept and the plan's features.
How accurate is ryser.id?
It is best on released music: a classic house or trance set usually comes back nearly complete. Heavy layering, edits, bootlegs and unreleased tracks often stay unknown. Every row says how sure it is, so you know which ones to check.
Why is a track I know marked unknown?
Usually because the DJ played it pitched, layered or as an edit, or because that version is in no database yet. A part of the set the recognition service did not answer for says Not scanned; run the set again to scan it. If you know the track, name it, and the correction teaches the next set that plays it.
What happens to my audio?
Audio from a link is processed and deleted right after the analysis. Your own uploads are kept 7 days, or for as long as you pay for a plan. Short fragments go to recognition services only to identify them.
Got a set link? Get its tracklist.
Paste the link and ryser.id names every track it can, with the time it starts. No account? The quick peek under the box does the first 30 minutes for free and mails you the list.
Other guides
Getting started with ryser.id: your first tracklistPaste a set link or upload a recording and get every track it can name, with times, BPM and key. With or without an account, what it costs, how to check the result and what to do with it.Updated 16 October 2026
Reading the setlist: the marks, the rows and checking every trackWhat Proven, Found, Weak, To check and Unknown mean, what each part of a row does on a phone and a computer, how Check every track works, and how to fix a track.Updated 14 October 2026
Minutes, plans and passes: what ryser.id costsEvery set finishes, and you only pay for what we find. Every account gets free minutes each month; plans add hours, paid by the month or once for life, and packs add minutes that never expire. What each plan gives, how the pass works, refunds and switching.Updated 16 October 2026