AudD for record labels and publishers

If you own or administer recordings, your day-to-day questions are about your catalog: is someone using our tracks without a license? Did an unreleased master leak? Is a new release sampling one of ours? Which stations are actually playing our artists? This page answers each of those questions in turn, with a recipe that already implements the workflow.

AudD’s music-industry users include Sony, Warner Music Group, and Universal Music Group. The mechanism underneath is a music-recognition HTTP API: you upload your catalog to a private fingerprint database on your account, then run audio — uploads, social URLs, radio streams — through recognition and get back which of your tracks appear, where, and how often.

What you can do

Find unauthorized use of your catalog across platforms

The core move for a rights owner is to flip recognition around: instead of “is there any commercial music here,” ask “is my recording here.” You do that by uploading your masters to a custom catalog (POST api.audd.io/upload/, special access required — email [email protected]). Once your tracks are in your account’s private fingerprint database, recognition calls can match incoming content against them.

When user-uploaded or published content reuses one of your recordings, recognition against your custom catalog returns an audio_id identifying the exact uploaded master that matched, so you know precisely which work was used.
The audio_id is also how you tell a catalog match apart from a match against the public 160-million-song database: it appears only on custom-catalog matches, and on those matches artist/title may be null because that metadata comes from what you uploaded.

Detect leaks of unreleased music

Unreleased masters are the highest-stakes case: the track isn’t in any public database, so only you can recognize it. Upload the unreleased recording to your custom catalog, then scan the places leaks surface.

Because an unreleased master lives only in your custom catalog, a recognition match against it is unambiguous — no public recording can produce a false positive for an unreleased work.
Leaks surface in long uploads, videos, and streams, not just clips. The enterprise endpoint (POST https://enterprise.audd.io/) chunks long audio and video server-side, and the streams surface watches continuous live audio; both can match against your custom catalog.

Detect samples and interpolations

A sample is a fragment of one of your recordings reused inside another track — often short, often pitched or time-stretched. Recognition against your custom catalog finds the reused segment and tells you where in the matched recording it came from.

Matching incoming content against your custom catalog returns the audio_id of the sampled master and a timecode — the position within your recording where the matched fragment sits.
To scan many candidate tracks rather than one, run your release-screening queue through the enterprise endpoint, which processes long files and full uploads against your catalog.

Track radio airplay for your artists

Royalty conversations and marketing decisions both need real airplay data: which stations play your artists, and how often.

The streams surface (addStream, setCallbackUrl, and GET /longpoll/) recognizes songs off many radio streams around the clock — an independent record of where your tracks are played on air.
To get airplay specifically for your catalog rather than a generic now-playing log, match the stream recognitions against your custom catalog, keep only the rows that are yours, and aggregate them into an airplay chart.

Where to start

Detect samples and reuse of your own tracks — the foundational recipe: upload your catalog, then identify catalog matches (samples, leaks, re-uploads) in incoming content via audio_id.
Monitor radio airplay for your music catalog — register many radio streams, record every recognition, filter to your catalog, and aggregate into an airplay chart.
Public database vs your custom catalog — the concept underneath all of the above: how a custom-catalog match differs from a public match, and how to tell them apart in a response.
Standard, enterprise, or streams: how to choose — which recognition surface fits each input: a clip, a long upload, or a live stream.

Identify a catalog match

Once your masters are uploaded, a custom-catalog match is identified by its audio_id. The shape below is the heart of leak and sample detection: a match that came from your catalog rather than the public database.

from audd import AudD

audd = AudD("your-api-token")  # token from dashboard.audd.io

# Recognize incoming content; matches can come from your custom
# catalog or the public 160-million-song database.
result = audd.recognize("https://audd.tech/example.mp3")

if result is None:
    print("no match")
elif result.audio_id is not None:
    # audio_id present => matched one of YOUR uploaded masters
    print(f"catalog match: audio_id={result.audio_id} at {result.timecode}")
else:
    # a public-database commercial recording
    print(f"public match: {result.artist} — {result.title}")

A non-null audio_id is the unambiguous signal that the audio matched a recording you uploaded — a leak, a sample, or an unauthorized reuse. On those matches artist and title reflect what you supplied at upload time and may be null if you didn’t supply them.

{
  "audio_id": 184237,
  "timecode": "00:14",
  "song_link": "https://lis.tn/Warriors"
}

Custom catalog upload needs special access. Uploading your masters to a private fingerprint database via POST api.audd.io/upload/ is enabled per account — email [email protected] to turn it on before you build the upload step.

Related