Applications

What an identifier hidden in a music file is for, and an honest note about what this system can and cannot do today.

Provenance

A mark that says where a recording came from. If you release music and want to know where a copy that turns up somewhere else was published first, the identifier is the thing that answers it.

This is the case the project is built for, and it is also the case with the hardest requirement: the mark has to survive being transcoded and re-encoded on the way, because that is exactly what happens to a file between publication and a listener.

Broadcast and podcast identifiers

A station identifier that travels with the audio rather than needing a separate stream marker. WebBeeps was built with this in mind, which is where the idea of a spoken or tonal identifier in a broadcast signal came from.

Linking a track back to a page

A short URL or IRI in the audio, so a listener who wants to know where a track came from can be told without being told separately. This is what the original site was for, and it is the smallest payload the inaudible scheme could carry if it became robust.

What this cannot do yet

Stated plainly, because a watermark that does not do what someone assumes is worse than one that is honest about its limits.

It has not been proven inaudible. The watermark is built to sit under the music and has been measured as a signal, and nobody has confirmed by listening that you cannot hear it on a given track. Play the marked file against the original, on a quiet passage as well as a loud one, before you publish it.

It survives a lot, and not everything. On the music it was measured on, it survived gain changes, dither, moderate noise, filtering, resampling, a crop, a fold to mono and MP3 at 128 and 64 kbit/s. It does not survive heavy noise, and results differ between tracks. It needs a track of at least 12 seconds, about 26 for a 22-character identifier and a minute for the longest, and a mark in a short clip has few copies to fall back on.

It is not a DRM and not an access control. Nothing here prevents anyone from removing a mark who wants to. A key stops anyone finding it without the key. It does not stop someone degrading the file until it no longer reads, and that has not been measured. It identifies a source; it does not enforce anything.

A mark made with the default key can be read by anyone. That is the right choice for an identifier meant to be looked up, and the wrong one for anything private. Give a key for anything that should not be found by someone who does not have it.

The audible scheme survives less than you would like, and you can hear it. It is the original WebBeeps scheme, kept so that old files still read. It is not a watermark for a finished track, and it replaces the audio with beeps rather than marking your music.

The marked file is a WAV. It keeps the original's bit depth if that was a WAV, and an MP3 becomes a 16-bit WAV. Formats other than WAV and MP3 are not read.