Mic Latency Test

What is this tool?

The Mic Latency Test measures an acoustic loopback: the page plays a short reference pulse through your speakers, listens for it on your microphone, and times the gap. That gap is the whole round trip — output buffering, the air between speaker and microphone, input buffering and the browser's own processing.

The page names the result precisely: Observed Acoustic Loopback Latency. It is deliberately not called microphone latency, because the microphone is only one segment of the path being timed.

Before measuring it calibrates. A two-second listen establishes your background noise floor, and the detection threshold is then set 14 dB above that floor and clamped to the range −50 to −15 dBFS. That is why the tool works in a quiet room and a noisy one without you adjusting anything.

How to use it

1. Use speakers, not headphonesThe test needs sound to travel through the air to your microphone. With headphones on, the microphone hears nothing and detection times out — the page explains this under “Why Headphones May Not Work for Acoustic Loopback”.
2. Allow microphone accessThe browser will prompt for permission. If you previously blocked this site, clear the permission from the address-bar icon and reload, because the prompt will not reappear on its own.
3. Let it calibrate in silenceCalibration listens for about two seconds and reports your noise floor as QUIET, NORMAL or HIGH — quiet is below about −55 dBFS, normal up to about −40 dBFS, and anything above that is high. Stay quiet for the full two seconds so the threshold it derives is correct.
4. Choose a reference frequency and set a moderate volumeReference pulses are available at 250 Hz, 500 Hz, 1 kHz, 2 kHz, 4 kHz and 8 kHz. Mid frequencies are usually detected most reliably; the level needs to be clearly above your noise floor without clipping the input.
5. Run the ten-round testA single measurement tells you very little. The ten-round test builds a distribution and reports average, median, minimum, maximum, P95 and jitter, which is what makes the result meaningful.
6. Run the stability test if the numbers move aroundThe stability test repeats measurements to show whether latency holds steady or drifts. Drift usually means a power-saving state, a busy CPU, or a wireless audio path.

What the results mean

Average versus median

The average is pulled around by a single bad round; the median is not. When the two disagree, the median is describing your typical latency and the gap between them is telling you that some rounds went wrong. The page discusses this under “Average vs Median Latency”.

Jitter

Jitter is the standard deviation of your measurements, reported to one decimal place. A low figure means a predictable pipeline; a high one means the delay itself is unstable, which is often more disruptive for monitoring and recording than a higher but steady latency. See the page's “Latency vs Jitter” section.

P95 and the maximum

P95 is the value 95% of your rounds came in under — the practical worst case, ignoring one freak outlier. The maximum is that outlier. If P95 sits close to the median, the pipeline is consistent.

Detection confidence

Each round carries a confidence derived from the signal-to-noise ratio of the detected pulse: below 6 dB is LOW, below 12 dB is MEDIUM, and above that is HIGH. LOW confidence rounds mean the pulse barely stood out from the room — raise the volume or reduce the noise rather than trusting the timing.

Measurement quality

The overall verdict is UNSTABLE when more than 30% of rounds failed to detect or jitter exceeds 18 ms, FAIR when jitter exceeds 8 ms or your noise floor was rated HIGH, and GOOD otherwise. It is a comment on the reliability of the run, not a grade for your hardware.

What it can detect

• The full round-trip delay of your current audio path, as observed rather than inferred
• Whether that delay is stable or drifts between rounds
• A high background noise floor, measured before testing begins
• Rounds where the pulse was too weak to detect confidently, flagged rather than silently included
• The latency penalty of a wireless output or input, by comparing runs

Limits worth knowing

A browser can only report what the platform gives it. These are the honest boundaries of this page, so a result is never read as more than it is.

• Nothing is uploaded or stored on a server. The engine contains no fetch, no XMLHttpRequest and no WebSocket; audio is analysed in the page and discarded. The tool does not save a recording of your microphone, which its own privacy FAQs state.
• It is not microphone hardware latency. The measurement necessarily includes output buffering, the acoustic travel time between speaker and microphone, input buffering and browser processing — the page names the figure Observed Acoustic Loopback Latency for exactly that reason, and has a “Does this test measure microphone hardware latency?” FAQ that says no.
• It cannot separate the segments of the path. If the number is high, the tool can tell you it is high but not which stage owns the delay; changing one variable at a time — wired instead of Bluetooth, a different microphone, a quieter room — is how you narrow it down.
• Detection has a 1,500 ms window. If the pulse is not heard in that time the round fails and reports that the microphone did not detect the reference audio signal, which is a detection failure, not a latency reading of 1,500 ms.
• It is not a human reaction-time test. Nothing here depends on you pressing anything: the pulse is generated and detected automatically, which the page contrasts with reaction timing under “Why Human Reaction Time Is Different”.

Tips

• Put the speaker and the microphone close together and keep them there for the whole run. Sound travels about 34 cm per millisecond, so moving them changes the number by a real amount.
• Test wired, then Bluetooth, with everything else identical. The difference between the two averages is the wireless penalty, and that comparison is the most useful thing this tool produces.
• If measurement quality comes back FAIR or UNSTABLE, fix the input conditions before re-reading the numbers — raise the pulse level, quieten the room, close heavy background applications — rather than averaging more bad rounds.

Troubleshooting

Related documentation

• AudioMicrophone TestLive level meter, waveform, spectrum and a local recording check.
• AudioBluetooth Audio TestCheck a paired Bluetooth output by ear: stereo, dropouts, bass, treble and a reaction-time delay check.
• AudioHeadphone TestEars, balance, separation, bass and treble checks built for headphones and earbuds.
• AudioSpeaker TestLeft, right, stereo, bass, treble and sweep tones for laptop or desktop speakers.
• AudioAudio Frequency TestA 20 Hz to 20 kHz tone generator with sweeps, steppers and A/B comparison.