About

NIL HEARD is not a band. It is a listening post. The record is assembled from shortwave signal that has already travelled — bounced off the ionosphere, lost information to fading, arrived slightly wrong — and this is the log of every reception that fed it. The claim on the sleeve is nothing was performed, everything was received. Here that claim is checkable.

Method

Every capture is made through a public KiwiSDR receiver, chosen for the path to the transmitter, recorded headless with kiwirecorder as 12 kHz mono WAV. The rig writes the provenance the moment the file is closed: UTC, frequency, mode, receiver location and antenna, the target's rights tier, and the sha256 of the audio. There is no separate bookkeeping step that can be skipped.

The rig's classifier cuts each capture into 20 s blocks and reads each three ways — spectral-peak prominence, keying rate and depth on the tone band, syllabic share of the envelope — and labels it carrier, keyed-tone, morse, voice, data or noise. The block timeline under every spectrogram is that output. It sorts material; it does not certify a station. That still takes ears.

This site is a build. It reads the rig's catalog — every reception, its provenance and what the classifier heard — and turns it into these pages and their JSON data; the Opus streams, the FLAC downloads and the spectrograms are made once by the rig from the audio it holds, and linked. No backend, no database, no accounts, no third-party scripts, no analytics. Same inputs, same site.

Rights

The recordings are ours: fixing a radio signal to a file is an act of recording, and the recording carries its own copyright. The underlying transmission is someone else's, and for anything not addressed to the public the law (German § 5 TTDSG, and its Swiss counterpart) says the content may not be passed on. So the archive publishes by tier:

• Tier A — time signals, VOLMET, NAVTEX, weather fax, RTTY, beacons, the empty band. Addressed to everyone. Published in full.
• Tier B markers and tones — buzzers, pips, interval signals, polytones, data bursts. A buzz is a sound, not a message. Published in full.
• Tier B voice or morse message content — spoken or keyed groups. Never verbatim: at most a 10 s excerpt from the start of the capture, or nothing.
• Tier C — anything private. Never recorded, never kept.

No transcripts, anywhere. Descriptions on this site describe sound — “keyed tone at 562 Hz, 0.3 Hz”, “voice, English, weather read on a loop” — never what was said. This is deliberate and it is the reason the archive can exist.

Receivers are credited by location and antenna as their operators list them publicly. The receiver's address is never published. Operators who prefer not to be named: one mail, withheld within a day.

What is listed, and what you can play

Everything. Every capture the rig makes is listed here and published in full — the recording to stream and download, its spectrogram, the receiver, the frequency, the UTC second, what was heard and the sha256 of the file — new ones as soon as the next build runs. There is no selection: near-identical takes of the same night stand side by side, because this is a log, not a portfolio.

Licence

Published captures: CC BY 4.0 — credit NIL HEARD / Reception Log (licence text). The code that made the record and this site is under the repository's code licence; the two are separate on purpose.

Automated collection of the audio for machine-learning training is not permitted; robots.txt says so to the crawlers that read it.

Contact

log@nilheard.com — takedown requests, receiver opt-outs, corrections. NIL HEARD declines interviews but answers correspondence.

Keys

space play/pause · ← → ±5 s · [ ] loop in/out · ↑ ↓ volume · m mute · j k next/previous capture in the list · #t=134 in a capture URL starts at 2:14.