Audio Latency & Timing-Offset Test

A live test that measures how far your taps land from the beat, on the audio clock.

What this test does

This tool plays a steady metronome click and asks you to tap on every click. It records the gap between each tap and the nearest click, then reports your average net offset in milliseconds, along with how much your taps scatter around that average.

Short answer: the net offset is how far ahead of or behind the click you tap, averaged over many taps. A negative number means you tap before the click; a positive number means you tap after it. Crucially, this figure is not a pure hardware reading. It bundles three things at once: the time your device takes to actually sound the click, your own reaction to what you hear, and the delay in registering your tap. Read as a personal calibration number, it is genuinely useful. Read as a lab measurement of your soundcard, it is not.

How to use it

Press Start and let a few clicks pass so you can lock onto the pulse. Then tap on every click, either with the on-screen button or by pressing the Space bar, and keep going without trying to correct yourself. The widget matches each tap to the nearest click and updates two figures live: your mean net offset and the standard deviation of your taps.

The numbers jump around at first and settle after a few dozen taps, so give it thirty to fifty before you trust the result. For a fair reading, use the same device, output, and headphones you actually play or game with, since each setup behaves differently. Run it two or three times and compare the results rather than reading too much into a single pass.

What it measures (and what it doesn't)

Your net offset is a sum, not a single quantity. It contains the audio output latency (the time between the software scheduling a click and your ears hearing it), your reaction and motor delay (hearing the click and moving your finger), and the input latency of the button or key being registered. Because those are folded together, the result is not a clean hardware-latency figure. It is your whole ear-to-finger loop on this device.

Simple auditory reaction time, responding to a sound the instant you hear it, typically runs around 150 to 250 ms. You might expect your taps to land that far behind each click. In practice they often do not, because tapping to a steady, predictable beat is anticipation, not reaction. When people synchronize taps to a regular pulse they tend to land slightly before each beat, a robust finding known as the negative mean asynchrony, a core effect in sensorimotor synchronization (Repp, 2005; Aschersleben, 2002). So your mean can easily come out negative. The standard deviation is a separate story: it measures your timing stability, how tightly your taps cluster, independent of whether they run early or late.

The math

For each tap the widget finds the closest click and computes:

offset = tap_time − nearest_click_time, measured on the audio clock.

Pairing every tap with its nearest click keeps each offset inside a window of half a beat either way, so a tap is never mis-assigned to the click before or after it. The mean of those offsets is your systematic offset: a consistent lead or lag. The standard deviation is your consistency: how far each tap wanders from your own average. The sign convention is fixed throughout: negative means you tapped before the click, positive means after. A large mean with a small deviation is a steady offset you can calibrate out; a large deviation is scatter you cannot.

Reading your result

Read the two numbers together. The mean tells you direction and bias; the standard deviation tells you how reliable that bias is. A tight deviation with a large mean is easy to calibrate around, because the offset is stable. A wide deviation means the underlying timing is noisy, and no single correction will fix it. Treat the bands below as rough guides, not verdicts, and compare against your own repeated runs rather than anyone else's number.

Why it matters

Most rhythm games ask you to set a timing offset so your taps line up with the music, and a stable net offset is exactly the number that setting wants. Measure it here, then dial it in and let the game feel honest. The same idea underlies monitoring latency in a DAW: when you play a live part through software, the round trip from input to output adds delay, and players compensate by pushing slightly ahead (MDN Web Docs, AudioContext.outputLatency). Knowing your own bias, including whether you naturally tap ahead of the beat, helps you separate a genuine equipment delay from your personal negative mean asynchrony, so you calibrate the machine without over-correcting yourself. If your deviation is high, steady it with silent-bar metronome drills and by learning to feel rushing and dragging; to find your natural pace instead, use the tap-tempo tool or read what BPM means.

Limitations

This is a personal calibration aid, not a hardware measurement. Because reaction, audio output, and input delay are conflated, the tool cannot tell you which part of the loop dominates. Wireless paths make it worse: Bluetooth headphones can add tens to over a hundred milliseconds of output latency, so a wired setup gives a cleaner reading. Browsers, operating systems, and devices each schedule audio differently, and the same person can get different numbers on different machines, or even different tabs. Small samples are unstable, so a handful of taps means little. Take several runs, keep the conditions consistent, and read the result as a ballpark for one setup rather than a fixed property of your hardware or your ears.

Sources: Repp, B. H. (2005), Sensorimotor synchronization: A review of the tapping literature, Psychonomic Bulletin & Review 12(6), 969–992; Aschersleben, G. (2002), Temporal control of movements in sensorimotor synchronization, Brain and Cognition 48(1), 66–79; MDN Web Docs, AudioContext.outputLatency.