Four real hydrophone recordings from NOAA, and two beds generated here. Which is which is written on every channel, because the difference matters.
The four recordings are real, and they are public domain. They come from NOAA's Pacific Marine Environmental Laboratory Acoustics Program. Works of the U.S. federal government are not under copyright, so no attribution is strictly required — it is given on every channel anyway, because it is the right thing to do and it costs nothing.
They are streamed from NOAA and never copied here. The audio is not proxied, cached or
re-served by this site; your browser fetches it from pmel.noaa.gov directly, which
is the same rule the song guesser's previews run under. There is still no audio file in this
repository.
This is the one drawer in the cabinet that plays a recording at all, and that is worth saying plainly. Everything else here is generated — oscillators and noise buffers, made at the moment you hear it. Real hydrophone data cannot be synthesised honestly: a made-up whale is not a whale. So this drawer carries real recordings and labels every channel RECORDING or GENERATED, and never lets you mistake one for the other.
Why the mixing is simple. NOAA's server sends no
Access-Control-Allow-Origin header, so these files can be played but cannot be read
into Web Audio — anything fancier than a level fader would silently produce silence. The
faders are therefore plain element volume, and the cabinet's mute switch is wired to them by
hand rather than through the shared bench.
The clips are short — between nine and thirty-five seconds — and loop. Two are recorded at a 5 kHz sample rate, so they only carry sound below about 2.5 kHz and will sound muffled. That is not a fault in the playback; that is what the hydrophone recorded.