A distributor receives an audio file.

How could it possibly know that part of the recording came from somebody else’s song?

One answer is audio fingerprinting.

What is audio fingerprinting?

A fingerprinting system creates a mathematical representation of characteristics within a recording.

That fingerprint can then be compared with reference recordings.

It is conceptually similar to recognising a fingerprint rather than comparing filenames.

Calling a file:

my-original-song-final.wav

does not change the audio inside it.

Exact copies are relatively straightforward

If somebody uploads the same recording again, recognition technology may identify the match.

But systems are becoming more sophisticated.

Altered recordings

Some technology can identify source material after changes including:

• speeding up;
• slowing down;
• pitch shifting;
• editing.

SoundOn, for example, announced technology from ACRCloud earlier this year specifically designed to identify derivative recordings despite substantial speed or pitch changes.

Does a match prove infringement?

No.

This is important.

There may be legitimate reasons recordings match.

For example:

• licensed samples;
• catalogue transfers;
• authorised remixes;
• public-domain material;
• properly licensed third-party audio.

A detection system identifies similarity.

Rights still require context.

Why human review matters

If every audio match automatically produced a rejection, legitimate releases would sometimes be blocked.

A distributor may therefore ask for:

• licence;
• ownership evidence;
• sample clearance;
• explanation.

Final thoughts

Audio recognition makes it increasingly difficult to hide copied recordings through simple technical changes.

The correct solution to copyrighted audio is not trying to evade detection.

It is obtaining the rights.