The friendly version

The five questions everyone asks.

Straight answers about what Duona does, how it works, and why a label or artist would care.

Question 01

What is an ISCC, and why does everyone in this space keep saying it?

A code a recording earns from its own sound. Nobody assigns it. Run the audio through the open math and the code comes out. Run the same file through it in Tokyo and the same code comes out there too.

Think of it as the difference between a face and a name tag. A name tag can be peeled off. A face just is. Strip the file's title, delete the credits, rename it to final_final_2.mp3: the sound part of its code stays the same, so the recording can still be recognized.

The reason lawyers and platforms take it seriously: it is an ISO standard, from the same standards family that runs the ISRC. Machines that have never heard of you can compute it, and when a registry holds your terms, find them sitting right next to it.

One honest line: the code says which recording this is. Whose it is, and what the rules are, come from the record we file beside it. Together they buy you a catalog machines can recognize without asking you first, which is exactly the audience you need when the machines are the ones doing the scraping.

Question 02

What is the difference between a fingerprint and a watermark?

Same jacket, two different tricks.

A fingerprint is recognizing the jacket on a stranger across the street. Nothing was ever added to it. It works on copies made before you ever showed up, because it works from what the thing already looks like. Cut the jacket up small enough, or re-tailor it hard enough, and recognition gets shaky: we publish exactly where.

A watermark, the stamp, is the name tag sewn inside the lining. You sew it in before the jacket leaves the house. It rides through the wash and most dye jobs, and when someone looks inside, it says whose it is and what the rules are. Some treatments do tear the tag out. We publish those too, on the page where you can try it.

One recognizes. One carries your note. Recognition finds copies you never touched. The stamp speaks for the copies you sent out yourself. The net and the signature.

Try both on one song, start to finish, on the stamp page. Two minutes, no account.

Question 03

What should a record label do with this technology?

Three moves, smallest first.

Move one: put a do-not-train record on the masters in your catalog. Since 2 August 2026 the European Commission can fine the companies that make general-purpose AI models when they ignore obligations that include respecting machine-readable rights reservations. That obligation sits on the model maker's side. Your side is that the reservation exists somewhere a machine can read it, and stays current. That is a filing plus upkeep, not a project. We file it and keep it current.

Move two: stamp what leaves the building. Promo copies, previews, masters headed to partners. In our tests the stamp took about a fifth of a second per song and stayed quieter than the hearing threshold in every moment we measured.

Move three: when a copy shows up somewhere it should not be, pull the receipt out of the audio and attach it to the takedown you were already going to send. Every takedown form asks which recording and says who. The receipt answers both, from inside the sound.

None of this changes how you work. It is a subscription-shaped habit that makes the catalog legible to machines and easier to defend with people.

Question 04

Okay, but what IS "this" technology?

Fair, because it is really three small things wearing one coat.

The code: identity computed from the sound itself. The face.

The record: a machine-readable note that says ours, do not train, filed where machines look. The land registry entry.

The stamp: your name and your terms riding inside the audio of copies you send out. The name tag sewn into the lining.

Mix them for different outcomes. The code covers recognizing what is already out in the world. Code plus record is the legal lever, the thing the EU fine hangs off. Add the stamp and your note travels inside the copies themselves, so even a stripped, renamed file walks around carrying your receipt.

The magic trick, and you can run it yourself right now: hand a roughed-up copy back to the reader and it tells you whose it is, whose terms it carries, and when it was stamped. That is the whole show. The rest is us keeping the paperwork current so you never think about it.

Question 05

Why would my artists care about this?

Because attribution is the whole game now, and the game is getting worse.

AI covers, sound-alike models, ripped re-uploads: streams that cannot be attributed to your artist cannot be monetized by your artist. The old lane markers, tags, titles, credits, are the first things a scraper deletes.

This puts the artist's name inside the sound, and their no-training terms on the public record. When a copy of their master shows up wearing somebody else's filename, it still answers to them.

It is also a story an artist can actually tell: my masters carry my name inside the audio, and there is a standing public record that says no model trains on me. That sentence does work in a press quote, in a negotiation, and on a phone call with a nervous manager.

And it costs like a plugin, not like a lawyer.

Stamp a song Talk to Duona