◷ Tempus

A toolkit for watch enthusiasts

Every mechanical watch
tells two stories.

The time on its dial, and the story of how well it keeps it. Tempus turns your phone into a timegrapher, a collection log, and a reference library — no account, no cloud, everything stays on your device.

01

Acoustic Timegrapher

Hold your watch to the microphone. Tempus listens to the escapement, detects the beat rate, and estimates how many seconds per day your watch gains or loses.

Start listening →
02

Collection Log

Keep a private record of every piece you own — reference numbers, movements, service history and photos. Stored locally, exportable as JSON whenever you want a backup.

Log a watch →
03

Reference Finder

Search a curated database of notable references, or snap a photo and let AI suggest the brand, model and reference number of a watch you've spotted in the wild.

Identify a watch →

Anatomy

Beneath the dial,
a heartbeat.

Twelve parts, one purpose.
Tempus keeps every one of them honest.

About this project

Tempus started as a simple question: can a phone microphone hear a watch escapement well enough to measure its rate? It turns out that in a quiet room, with the caseback pressed gently against the mic, it can. A mechanical watch ticks at a fixed beat rate — most modern movements run at 28,800 beats per hour, vintage pieces often at 18,000 — and each tick is a sharp acoustic transient the Web Audio API can pick out.

Measure enough tick intervals, compare them to the ideal interval for the detected beat rate, and you get a rate deviation in seconds per day — the same figure a watchmaker's timegrapher reports. It won't replace a proper machine (there's no amplitude or beat-error reading from a phone mic), but it's a surprisingly honest first opinion on whether your watch needs a service.

The rest of the app grew around that idea: somewhere to log the watches being tested, and a reference library for identifying pieces. It's a static site — plain HTML, CSS and JavaScript, no framework, no backend. Your collection never leaves your browser.

Acoustic Timegrapher

Find a quiet room. Rest the caseback of your watch directly on the phone, near the microphone, and keep still while it listens.

Ready — press start when the watch is on the mic.
Rate — sec / day
Detected beat — bph
Beats tracked 0 across the recording
Signal
no input
How the measurement works — and its limits

Tempus doesn't try to hear each tick on its own — on a phone mic they're often buried in noise. Instead it records the high-frequency energy envelope, finds which standard beat rate the audio is periodic at using autocorrelation, refines the period with a ladder of longer and longer correlation lags, locates the beat phase, then picks the loudest moment inside a narrow window around every predicted beat. A straight-line fit through those beat times gives the true period — and the rate in seconds per day, with a ± uncertainty from the fit residuals. A phone mic can't measure amplitude or beat error, and heavy noise or movement can still prevent a lock — but the beat-pattern approach works well below the point where individual ticks are audible. Quartz watches tick once per second and are usually still too quiet to measure.

My Collection

No watches logged yet.

◷

Your collection lives here — and only here.
Everything is stored in this browser, never uploaded.

Reference Finder

Search the reference library, or identify a watch from a photo.

⌖

Take or choose a photo of the watch.
Dial-side, in decent light, works best.

API key settings

Photo identification sends the image to the Claude API using a key you provide. The key is stored only in this browser. Get one at console.anthropic.com.