A native macOS app that watches a microphone for noise in a frequency band you choose, records timestamped evidence of it, and exports a report you can hand to a landlord, council, or mediator.
- Live frequency analysis — a real-time spectrum and a band-level meter, with Fast/Slow time weighting.
- Drag to set the band and threshold — target exactly the frequencies you care about (e.g. a subwoofer's 20–120 Hz).
- Event detection — automatically logs each time the band crosses your threshold, with peak/mean level, duration, and timestamp.
- Recording — saves a short band-filtered clip per event, with pre-roll.
- Proven coverage — records exactly when monitoring was running, so the report can prove "we were listening", not merely assume it.
- Review — a calendar heatmap, a 24-hour event ribbon, an event list, and clip playback.
- Evidence export — a paginated PDF report, CSVs, and the in-range clips, zipped into one bundle.
Bandwatch documents the pattern of noise in a chosen frequency band — when events happen, how often, how long they last, and their relative loudness over time, each stamped with the input device that captured it. It is a monitoring log, not a certified acoustic instrument:
- Levels are dBFS, not calibrated SPL — a relative digital scale, not a decibel reading you could cite as an absolute loudness. Establishing an absolute level (e.g. against a noise ordinance) needs a calibrated sound-level meter or a professional.
- The equipment is not calibrated — a consumer microphone/interface with uncharacterized sensitivity and low-frequency response (see Choosing a Microphone below).
- The detection threshold is set by you — an "event" is the in-band level crossing the threshold in effect at that moment, and each event stores the exact threshold that applied.
- Recorded audio is band-filtered — only frequencies inside the chosen band are kept. For a low bass band (e.g. a subwoofer's 20–120 Hz) this strips out the speech range, so those clips contain no intelligible speech. A band that overlaps voice frequencies (roughly 300 Hz–3.4 kHz — including the 1–4 kHz and 2–8 kHz presets) can capture clear, understandable speech, so treat bundles from higher bands as potentially containing conversation before sharing them.
- Coverage is a measured record of when the recorder was running, not an inference.
- Bandwatch records short per-event clips, not continuous audio — only the moments that cross your threshold are kept.
- Bandwatch never deletes your recordings. If free space drops below a warning level (default 16 GB), it warns you with an in-app banner so you can act before recording is affected. If free space keeps falling and crosses a hard floor (default 8 GB), it stops recording and logs a gap rather than writing onto a full disk, or deleting existing evidence to make room. These defaults are conservative placeholders, not tuned to measured disk usage — back up your data if you are relying on it.
The MacBook's built-in microphone is not reliable for band-filtered monitoring. Its aggressive noise suppression and voice-tuned processing distort low-frequency content, so a real subwoofer rumble can read faint or inconsistent. Use a discrete microphone — a USB mic, a lavalier, or an audio interface — for trustworthy measurements. Pick it from the Input menu on the main window.
Prefer a ready-to-run app? Each release
ships a prebuilt Bandwatch.app inside Bandwatch-<version>-universal.zip — a
universal (Apple Silicon + Intel), ad-hoc-signed build. Download it, unzip, and
follow the one-time Gatekeeper step in DEPLOY.md; no toolchain
needed.
To build from source instead, you need macOS 15+ and a Swift 6.3+ toolchain. No third-party dependencies — only Apple's system frameworks.
swift build
swift run Bandwatch(This repository uses a swiftly-managed toolchain; if swift on your PATH is
older, activate your toolchain first.)
MIT — see LICENSE. © 2026 Chris Connar.