A LITTLE CONTEXT GOES A LONG WAY
Audio Latency Test: Check Your Beat Timing
An audio latency test can help you explore how consistently you follow a sound. This browser beat check, also called an audio delay test, compares your taps with a scheduled rhythm. You get a human synchronization estimate, influenced by anticipation, your input device, and browser timing.
How to Use the Audio Latency Test
A few quiet beats. One consistent input method.
Find a comfortable beat
Choose your playback device in system settings. Start at 100 BPM and keep system volume low. Use Play preview to check the click before beginning.
Listen, then tap
Press Start and listen to four warm-up clicks. Tap the pad or press Space once per beat. Follow what you hear; there is no visual beat cue to chase.
Read a whole segment
Collect at least ten valid taps. The median and spread then appear. Stop keeps that segment visible. Pause clears it; Resume begins a new warm-up.
There is no fixed correct number for every headphone. Do not tap faster to push the result toward zero. This is an observation of your synchronization, not a score or a certificate for your playback equipment.
What Your Timing Result Actually Means
The main number is a signed tap phase offset: the middle, or median, timing difference between valid taps and their nearest scheduled beats, mapped to the browser's output timeline. Positive means late relative to that reference; negative means early. Values are rounded to whole milliseconds for readability, without implying one-millisecond physical accuracy.
At 120 BPM, beats are 500 ms apart. A tap 40 ms after its reference is +40 ms. A tap 25 ms before it is −25 ms. A tap 300 ms after one beat is also 200 ms before the next: its nearest-beat phase is −200 ms.
That last example explains modulo ambiguity. Phase wraps into the interval from minus half a beat up to, but excluding, plus half a beat. It cannot tell whether you followed a previous click or anticipated a later one. Delay spanning a whole beat is indistinguishable from no extra whole-beat delay in this method.
Why a negative result is possible
People can anticipate a repeating rhythm. Your finger may move before the scheduled reference even when the physical audio pipeline has positive delay. A negative phase is normal; it does not mean a device delivers sound before receiving it. The test never subtracts a guessed human reaction time to manufacture a hardware result.
Spread is the median absolute deviation, or MAD: a robust description of how far taps typically fall from their circular median. For offsets of 20, 25, and 30 ms, the median is 25 ms and MAD is 5 ms. This three-value example explains the statistic; the live result still requires ten taps.
Choosing Beats, Volume, and Input Method
The tempo range is 40–240 BPM, with a default of 100. A useful starting range is 80–120 BPM. The beat interval is 60,000 divided by BPM, in milliseconds. Change tempo while stopped; a new tempo clears the sample set so different beat lengths are never mixed.
The volume slider starts at 20% of a capped output gain. This cannot control your operating system's amplifier or guarantee a safe listening level. Lower system volume first, especially with earbuds. Muting the slider does not stop the scheduled rhythm, so use Stop when finished.
For comparisons, use the same keyboard key or the same touchscreen throughout. Space ignores held-key repeats and does not take over text fields, sliders, links, or other buttons. The tap pad responds to pointer press rather than release. Very close double taps and repeated inputs for one beat are rejected.
When an Audio Delay Test Helps
Try several short runs before and after changing from laptop speakers to wired headphones or Bluetooth earbuds. Keep the browser, tempo, input method, posture, and surroundings consistent. Write down the playback path, median, spread, and whether the segment was stable. This page does not save a device history.
Casual music practice and rhythm game calibration may benefit from seeing a repeatable early or late tendency. A game's sign convention may differ, and its own calibration process remains the relevant reference. This page cannot calculate a universal game offset or prove which headset is better.
Understanding Differences Between Sessions
The display uses a rolling window of up to 32 valid taps after warm-up. A stable segment means the taps cluster reasonably well under this tool's rules, not that the hardware measurement is accurate. Wide spread, a changing tendency, missed beats, or a result near the half-beat boundary produces an unstable segment.
A missed beat keeps the segment marked unstable until you start again. Switching tabs, interrupted scheduling, an unstable output clock, or suspended audio clears the segment. Pause also clears samples, because joining observations across an interruption would hide changed conditions. Stop retains the completed values until Start or Reset.
Switch playback devices only between runs. The browser can report some route changes but cannot expose every system-level switch. After any change, stop and begin again yourself. Noise cancellation, wireless routing, background load, practice, and fatigue can all affect the comparison.
What This Test Cannot Verify
No exact Bluetooth codec latency. Taps combine human prediction, sensorimotor synchronization, input processing, and audio delivery. A quiet, repeatable result still does not isolate any one component.
No acoustic arrival or round trip. This page does not request microphone permission or record sound. A calibrated external reference or suitable loopback setup is needed to investigate physical playback delay. Sound-and-picture alignment is another task; this beat check does not measure video sync.
No universal browser support. Web Audio must start after your action. Signed phase also requires usable output timestamps. Unsupported, stale, or unstable mapping leaves phase unavailable, even if clicks can play.
The optional diagnostics show browser-reported base and output latency estimates separately. They are not microphone-verified measurements, and a zero is displayed as reported zero rather than proof of an instantaneous device. The output timestamp documentation and output latency documentation explain these browser APIs.
Common Questions About Beat Timing
Do I need to allow microphone access?
No. This beat test uses playback and your taps only. No recording, audio upload, account, or microphone permission is part of this workflow.
Why can I hear nothing?
Press Play preview, then check system volume, the browser's site sound setting, and whether the tab is muted. Confirm the system output is the headset or speaker you intended. If audio could not load, check your connection and retry. On mobile, return to the foreground and start again after interruptions.
Why does the number never settle?
Try a slower tempo and ten or more deliberate taps after warm-up. Avoid alternating touch and keyboard. If you miss a beat, restart. When clock mapping is unavailable, waiting for more taps will not create a valid result; another browser may provide usable timestamps.
Does +50 ms mean my headphones add 50 ms?
No. It means your taps centered about 50 ms after the nearest browser output reference in that segment. Your anticipation and input path remain part of the observation. Compare several runs before interpreting any difference.
Should I use speakers or wireless headphones?
Use the playback path you want to observe, in a quiet setting and at a comfortable volume. Speakers introduce room sound and headphones change what you hear. Keep each setup's notes separate and avoid treating their tap offsets as technical specifications.
Method and Data Notes
Audio Delay Test provides this tool and guide to make manual beat observations understandable. The click is an original supplied PCM asset. Scheduled source starts use one Web Audio context; input timestamps are compared through validated output timestamps. The test holds its samples in this tab's memory, clears them on reload, and does not submit them to an application backend. No author credentials or laboratory validation are claimed.
AI assisted the implementation and explanatory copy. The worked examples describe calculations, not hardware trials. We use Google Analytics and Microsoft Clarity to understand website visits and page interactions. The Privacy page explains these services and how they relate to locally calculated test samples. The site has no advertising placements.