Audio latency test

A click out of the speaker, recorded by the microphone. The gap is your real delay, Bluetooth included.

What your browser declares

Audio engine base latency: not reported
Estimated output latency: not reported
Sample rate: not reported

These are values the browser reports, not measured by this page. They describe the software path only: Bluetooth headphones add their delay after this point and it does not show up here.

Measure it properly, with the microphone

Plays a click through the speakers and records it with the microphone. The difference is the full round trip, which is the only figure that includes Bluetooth delay.

It has to play through speakers, not headphones: the microphone needs to hear the click. Turn the volume up and do not cover the mic.

0 trials

Sync check

If you are wearing headphones the microphone cannot hear them and the method above will not work. This is the honest alternative: the circle lights up at exactly the moment the click sounds. If you hear the sound after seeing the flash, that gap is your headphone delay. It does not produce a number, and there is no way for it to.

Generated right here, in your browser. What you type never leaves your device.

Keep going with Audio test Speaker test Microphone test Fix my speaker

Audio latency is the time between the computer deciding something should sound and that sound reaching your ear. Over wired speakers it is a few milliseconds and nobody notices. Over Bluetooth headphones it can be two hundred, and then it shows up everywhere: lips in a video stop matching, a game answers late, and playing an instrument becomes impossible.

Almost every page on the subject discusses latency without measuring it. This one measures it: it plays a click through the speakers, records it with the microphone, and the gap between those two moments is the full round trip. It is the only method available to a browser that includes wireless delay, because the sound genuinely has to leave and come back.

You will also see what your browser declares on its own, which is useful but measures something else: the software path only, before the signal ever leaves the machine. And there is one case that cannot be measured at all — headphones you are wearing — for which there is an honest alternative below that does not invent a number.

Manual

How to use it

1
Play through speakers, not headphones

The microphone has to hear the click. Send it to headphones on your head and there is nothing to record, so there is no measurement to be had.
2
Turn the volume past halfway

The click is deliberately short so it is easy to locate. Too quiet and it blends into the room, and the page will tell you it did not hear it.
3
Grant microphone permission

Without it you only see what the browser declares. The audio never leaves your machine: it is analysed and discarded on the spot.
4
Let it run several trials

One is not a measurement. The page repeats up to twelve times and takes the median, which is what survives one odd trial without being dragged off.
5
Wait until it says the measurement has settled

It needs at least five trials that agree with each other. A wide spread means something is changing between trials and the number is not worth quoting yet.
6
Wearing headphones? Use the sync check

The circle lights at the exact moment the click sounds. What you perceive between the flash and the sound is the delay, even without a number.

In depth

Things worth keeping in mind

What can be measured here and what cannot

Worth being clear about before drawing conclusions from a number, because the three figures on this page measure different things.

Measured: the full round trip through the air, with speakers. Includes the Bluetooth delay of wireless speakers.
Read, not measured: the base and output latency the browser declares. Software path only.
Cannot be measured: headphones you are wearing. The microphone cannot hear them. Use the sync check.
Cannot be separated: how much of the total is output and how much is microphone input, beyond the browser-declared share.
Not attempted: a number derived from claps or key presses. That would be your reaction time dressed up as latency.

History

A problem digitisation brought with it

The analogue chain did not have this problem. A microphone into an amplifier into a speaker reproduces sound at the speed of electricity; the delay is however long the air takes to reach your ear, and nothing else.

Digital processing changed all of it. To do anything to sound you must convert it to numbers, and to work with numbers you must gather them into blocks. That block, the buffer, is the original source of computer audio latency: the smaller it is, the less delay and the greater the risk that the machine misses its deadline and you hear a click.

The music industry was the first to fight this seriously. Steinberg published ASIO in 1997 precisely to bypass the Windows audio layers, which added tens of milliseconds and made it impossible to record while playing along to an existing track. The threshold quoted ever since is around 10 milliseconds: below that, a musician does not perceive the gap between pressing and hearing.

Bluetooth brought the problem back at scale. The A2DP profile, from 2003, was designed for listening to music, where two hundred milliseconds of delay bothers nobody because there is nothing to compare against. The trouble appeared when those same headphones were used for video and for games, which is when the brain has a visual reference and spots the mismatch immediately.

10 ms: Threshold below which a musician stops noticing
1997: ASIO bypasses the Windows audio layer
2003: A2DP: Bluetooth built for music, not video
150-200 ms: Typical delay of the standard SBC codec

How it works

Why Bluetooth arrives late

The delay on wireless headphones is not the flight time of the signal through the air, which is negligible. It is the processing at both ends.

Audio does not travel raw: it is compressed before sending and decompressed on arrival. Compressing means gathering a block of samples before you can begin, and that block is a delay you never get back. On top of that the receiver keeps a cushion of sound in hand so a radio dropout does not come through as silence, and that cushion is more delay. Added up, the standard SBC codec lands around 150 to 200 milliseconds.

Low-latency codecs attack exactly those two things: smaller blocks and shorter cushions, at the cost of being more fragile in the face of interference. AptX Low Latency aimed at roughly 40 milliseconds, and the game modes on many recent headphones do the same. That is why the same headphones can behave on a phone and misbehave on a laptop: it is not the hardware, it is which codec the two ends negotiated.

Measuring that from a browser can only be done one way: by making the sound genuinely leave and come back. Any figure a browser hands you without using the microphone describes only its own software path, and falls short in precisely the part that is bothering you.

By hand

Measuring the delay with a phone camera

If you want a figure for headphones you are wearing — the case this page cannot measure — a phone and a video will get you one.

1

Position the phone so it sees both the flash from the sync check and, very close by, one of the earpieces.
2

At 240 frames per second each frame is about 4 milliseconds, which is far more resolution than you need here.
3

Step through frame by frame to the first one where the circle is lit. Write down the number.
4

In the video’s own audio track, locate the peak of the click. The difference in frames, times 4 milliseconds, is your delay.

Diagnosis

If something is not working

Symptom:It says it did not hear the click

Likely cause:: Either the sound is going to headphones, or the volume is too low, or something is covering the microphone.
What to do:: Check in the system that output is the speakers rather than a connected headset. Push the volume past halfway. On a laptop, make sure your hand is not resting over the microphone area.

Symptom:Every trial gives a very different number

Likely cause:: Nearly always system processing: echo cancellation or noise suppression, altering the click differently each time.
What to do:: Look in the system sound settings for microphone enhancements and switch them off. Close video-call applications too — they claim the microphone and apply their own processing even in the background.

Symptom:It measures nothing with Bluetooth headphones

Likely cause:: Expected, and there is no fix: the microphone cannot hear a sound playing inside your ear.
What to do:: Use the sync check below, which is the honest alternative. If you need an actual figure, film the flash and the earpiece together in slow motion on a phone, as the by-hand section explains.

Symptom:The browser reports no output latency

Likely cause:: Not all of them expose it. Firefox and Safari commonly leave it unreported; Chrome supplies it on most systems.
What to do:: Nothing to fix: the microphone measurement does not depend on that value and works the same without it. All you lose is the breakdown between the declared share and the remainder.

Symptom:The figure is higher than I expected on wired speakers

Likely cause:: The round-trip number also contains the microphone’s input path, which can be as slow as the output one.
What to do:: That is not a fault, it is what "round trip" means. To see the output side alone, read the breakdown, and bear in mind that only the browser-declared share can be attributed with confidence.

Symptom:It works less well on a phone

Likely cause:: Many phones apply audio processing that a browser cannot switch off, and the speaker and microphone sit very close together.
What to do:: Try a computer if you can. On a phone, the sync check is more trustworthy than the microphone measurement, because it does not depend on the click surviving that processing.

Trivia

Things worth knowing

Air adds about 3 milliseconds per metre

Sound travels at roughly 343 metres per second. Sit a metre from the speaker and those 3 milliseconds are already inside any round-trip measurement, and there is nothing to fix about them.

Video tolerates late sound far better than early sound

Broadcast recommendations allow audio to arrive well after the picture and almost not at all before it. There is a physical reason: in the real world sound always arrives after light, never before, and the brain is trained on that.

Your headphones do not have a fixed latency

It depends on the codec they negotiate with each device. The same headphones can run at 40 milliseconds on a phone that supports a low-latency codec and 200 on a laptop that only speaks SBC. The blame is shared between both ends.

Game modes cut latency by making the sound worse

There is no magic to it: they shrink the block size and the safety cushion. That trims the delay and in exchange makes the link more fragile, so somewhere with crowded Wi-Fi you are more likely to hear dropouts.

The buffer is a choice, not a fault

Smaller means less latency and a higher chance the machine misses its deadline and you hear a click. Audio software lets you pick it precisely because the right point depends on what you are doing.

Questions

Frequently asked questions

How much audio latency is acceptable?

Entirely dependent on what for. For listening to music it does not matter: two hundred milliseconds bothers nobody because there is nothing to compare against. For video, past about a hundred milliseconds you start noticing the lips do not match. For gaming, above about fifty it gets in the way. And for playing an instrument the threshold is around ten, which is where a musician stops perceiving the gap between pressing the key and hearing the note.

Why are my Bluetooth headphones so far behind?

Because the audio is compressed before sending and decompressed on arrival, and compressing forces you to gather a block of samples before you can start. On top of that the receiver holds a cushion in advance so interference does not come through as silence. Both add up, and on the standard SBC codec that is roughly 150 to 200 milliseconds. It is not the distance, and it is not the quality of your headphones.

Can I measure the latency of headphones I am wearing?

Not with this page, and it is worth saying plainly: the microphone cannot hear a sound playing inside your ear, so the round-trip method has nothing to record. The honest alternative is the sync check, which flashes at exactly the moment the click sounds and leaves you to judge the gap. It does not give a number, and no page using the browser alone can give you one. If you need the figure, the by-hand section explains how to get it with a phone’s slow-motion camera.

What exactly does the round-trip number measure?

The whole path: the browser prepares the sound, the system hands it to the card, it leaves the speaker, travels through the air, enters the microphone, crosses the input path and reaches the browser again. That is why it includes Bluetooth delay, and also why it is larger than the output side alone. Do not read it as "how long my speakers take": it is the sum of out and back.

Why not measure by asking me to clap in time?

Because that number would be the sum of the audio delay and your reaction time, and nothing separates the two. Reaction to a sound sits around a hundred and fifty milliseconds and varies with tiredness and attention, so we would be presenting as latency something that is substantially you. We would rather withhold the figure than publish an invented one.

Why is the figure the browser declares so low?

Because it describes only its own software path: how long from generating the sample to handing it to the system. Everything after that — the sound card, the speaker, the wireless link — falls outside. It is a real and useful number for comparing browsers, but it does not answer the question that brought you here if your problem is a pair of headphones.

Is my audio stored or uploaded?

No. The recording is analysed in your browser to locate the click and discarded immediately. Nothing is stored, nothing is uploaded, and the microphone permission is released as soon as you stop measuring.

Can I reduce my headphone latency?

Sometimes. If your headphones have a game or low-latency mode, turn it on: it shrinks the block size in exchange for more sensitivity to interference. Check which codec is being negotiated too, since it depends on both ends and an old Bluetooth adapter may be forcing SBC. And if what you need is to play an instrument or game seriously, the realistic answer is a cable: no wireless optimisation gets below what a jack does.