Microphone test

Watch the waveform and level as you speak, then record five seconds to hear yourself the way everyone else does.

Your browser will ask for permission. The audio stays on your machine: nothing is sent to a server and nothing survives a reload.

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

Keep going with Audio test Speaker test Audio latency test Decibel meter

This test opens your microphone from the browser and draws what comes in: the waveform, the current level and the peak. That tells you within seconds whether the microphone is picking anything up, and whether it is doing so at the right volume.

The audio never leaves your machine. Nothing is sent to a server, nothing is stored, and it all disappears when you reload. The test recording lives in the browser’s memory and is discarded when you close the tab.

The part most people skip is listening back. The meter tells you signal is arriving, but it cannot tell you there is a hum, an echo, or that your computer fan is sitting on top of your voice. Recording five seconds and playing it back is the only way to hear what the person on the other end hears.

Manual

How to use it

1
Give the browser permission

Press the button and accept when the browser asks. The permission covers this tab only and is dropped when you close it. If you blocked it out of habit, there is a fix further down.
2
Pick the right microphone

If you have several — the laptop one, a headset, a webcam — a picker appears. This is the most common fault of all: the system is listening through a different microphone from the one you think.
3
Speak normally and watch the level

The bar should move freely without hitting the end. Barely moving means the gain is low or you are too far away. Turning red means it is clipping, and your voice will sound distorted no matter how quietly you talk afterwards.
4
Read the spectrum

The coloured bars split the sound by frequency: bass on the left, treble on the right. Your voice should mostly fill the middle. A narrow spike hugging the left edge that stays put when you go quiet is electrical hum, not you.
5
Check the noise floor too

Stay quiet for a few seconds and watch that figure. It is what your microphone picks up when nobody is talking: fans, air conditioning, the street. The lower it is, the cleaner your voice will sound.
6
Record and listen back

Five seconds of talking normally. Playing it back gives you exactly what the far end gets, including the problems no meter can show.

In depth

Things worth keeping in mind

Condenser

This is what nearly every USB, laptop and phone microphone uses. A very thin diaphragm vibrates in front of a fixed plate, and that changing distance changes the electrical capacitance between them. It needs power to work, which over USB arrives on the cable itself.

They are very sensitive, and that is both the virtue and the problem: they catch the detail in a voice, but also the keyboard, the fan, and the conversation in the next room.

Very sensitive, good at picking up a voice at moderate distance
Take in a lot of room noise: better in treated or quiet spaces
Need power (USB, battery, or 48 V phantom)

Dynamic

It works like a loudspeaker in reverse: a coil attached to a diaphragm moves inside a magnet and generates current. It needs no power and handles very loud sources without distorting, which is why it is the live-stage microphone.

It is considerably less sensitive than a condenser, and that is its advantage for recording at home: you have to get close, so it captures your voice and very little of the room.

Ignore much of the room noise if you speak close to them
No power needed, and very hard to damage
Usually want more gain: on some interfaces they sound quiet

MEMS, the one in your phone

A miniaturised condenser etched onto silicon, about the size of a grain of rice. It is what phones, laptops and wireless earbuds use, which is why any modern handset has three or four of them.

With several, the device can compare what reaches each one and work out which direction a sound came from — that is how background noise gets cancelled on a call.

Tiny and cheap, surprisingly good for speech
Almost always paired with system-level processing on top
Degrade with humidity and dust over the years

Spectrum and spectrogram: what each one watches

The spectrum splits this instant of sound by frequency: it answers "what is sounding right now". The spectrogram stacks those instants one behind another, adding the dimension the spectrum lacks — time.

That difference is what settles the hard diagnoses. Mains hum and a held note look identical on a spectrum: both are a peak at one frequency. On the spectrogram the hum is a straight line that has been there since you opened the page, and the note is a stroke that starts and stops.

Continuous horizontal line: constant noise — a fan, electrical hum
Blobs rising and falling in the middle: normal speech
The whole upper half lit with nobody talking: hiss, gain set too high
Evenly spaced parallel lines: harmonics, characteristic of a voice or an instrument

Polar pattern: where it listens

A microphone does not hear equally in all directions. A cardioid pattern takes what is in front and rejects what is behind, and it is the one you want for talking. An omnidirectional pattern takes everything equally, which suits a meeting around a table and ruins a room with echo.

If your voice sounds distant even though the level is fine, it is often not the gain: you are talking into the wrong side. On many USB models the front is the side with the brand on it, not the tip.

History

From Berliner’s coil to the chip in your phone

The carbon microphone that Emile Berliner and David Edward Hughes developed separately around 1877 is what made the telephone practical. Carbon granules changed resistance as sound compressed them. It sounded poor and it hissed, but it was cheap and tough, and it stayed inside telephones for the best part of a century.

In 1916 E. C. Wente, at Western Electric’s labs, built the first condenser microphone. It was an enormous leap in fidelity, and it is the principle the microphone in your laptop still uses more than a century later.

The moving-coil dynamic arrived in 1931, also at Western Electric, bringing the ruggedness that let microphones leave the studio. The Shure SM58 of 1966 took that design to a point so thoroughly solved that it is still made almost unchanged today.

The decisive change for the device in your hand was MEMS, which reached the mass market in the early 2000s. Etching the microphone directly onto silicon made it possible to put several in a phone for pennies, and that is why any handset today can cancel noise by comparing what each one hears.

1877: Carbon microphone
1916: First condenser
1966: Shure SM58
48 kHz: Common sample rate today

How it works

How your voice reaches the browser

Your voice is a variation in air pressure. The microphone turns it into a variation in voltage — that is the analogue signal — and from there everything that happens is counting.

The analogue-to-digital converter measures that voltage many times a second. The standard rate for speech and music is 44,100 or 48,000 samples per second, and the number is not arbitrary: reconstructing a wave requires sampling it at twice its highest frequency, and human hearing reaches about 20,000 Hz.

Each sample is stored to a certain precision. At 16 bits there are 65,536 possible amplitude values; at 24 bits, over sixteen million. More bits does not mean more volume: it means more room between the circuit’s own noise floor and the point where the signal clips.

And clipping is exactly that: asking the converter for a value higher than it can represent. The wave gets sheared flat at the top, and that shearing is irreversible. Turning the volume down afterwards does not fix it, because the information that got cut is gone. That is why the meter warns before you reach the end.

The browser receives those samples and hands the page a list of numbers between -1 and 1. Everything this tool does is read that list: the waveform is drawing it directly, and the level is the root mean square of each block, which tracks perceived loudness better than the peak value does.

By hand

Tracking down background noise with nothing installed

If the meter will not drop below a certain level when you go quiet, something is making noise. You find it by elimination, and it usually takes under two minutes.

1

Let the noise floor settle for a few seconds and keep that figure as your reference.
2

If the number drops, the culprit is a fan or a drive. A mic sitting on the desk picks up vibration through the furniture, not just through the air.
3

A constant low hum is usually a ground loop through the power supply. If it vanishes on battery, you have found it.
4

They are broadband noise, the hardest kind to remove afterwards, and almost nobody notices them until they hear them recorded.
5

Repeat the five-second recording after each change. Comparing two recordings tells you more than watching the meter.

Diagnosis

If something is not working

Symptom:I blocked the permission by accident and it never asks again

Likely cause:: The browser remembers that decision per site and will not ask again while it is stored.
What to do:: Click the icon to the left of the address — a padlock, a slider, or a crossed-out microphone — find the microphone permission and set it to "Allow", then reload. On mobile it may also be blocked for the whole browser, in the system app settings.

Symptom:Permission is granted but no sound comes in

Likely cause:: Almost always a different microphone is selected from the one you are using, or the one you are using is hardware-muted.
What to do:: Switch device in the picker above. Then check the physical switch: many headsets have one on the cable and many USB microphones have a mute button with an LED. Finally check the system mixer, where the input volume may be at zero.

Symptom:I sound very quiet even when I shout

Likely cause:: Input gain is low, or it is a dynamic microphone plugged into an input that cannot drive it.
What to do:: Raise the input volume in your system sound settings. On Windows there is also a boost checkbox in the microphone properties. And get closer: most voice microphones are designed for a hand’s width away, not half a metre.

Symptom:I sound distorted or metallic

Likely cause:: You are clipping the input, or the system’s automatic processing is working too hard.
What to do:: Lower the gain until the meter stops hitting the end, and back off slightly. If it still sounds artificial, switch off noise cancellation and echo suppression in your sound settings: they try to clean the signal and frequently wreck the voice.

Symptom:There is a constant hum in the background

Likely cause:: Usually a ground loop through the power supply, or interference from a cable running near mains.
What to do:: Unplug the laptop charger and see whether it disappears. Move the microphone cable away from power cables. If it is USB, try another port, ideally on the other side of the machine.

Symptom:It works here but not in Zoom, Teams or Discord

Likely cause:: Each application has its own device selection, separate from the system’s and from the browser’s.
What to do:: Open that application’s audio settings and choose the microphone there. Working on this page rules out the hardware and the system permission: the problem is inside that application.

Symptom:The other person hears an echo of themselves

Likely cause:: Your microphone is picking up what comes out of your speakers and sending it back.
What to do:: Use headphones: it solves echo instantly and at the source. If you cannot, turn the speakers down and move the microphone away from them, and leave the video call app’s echo cancellation on.

Trivia

Things worth knowing

Your recorded voice sounds wrong for a physical reason

When you speak you hear two things at once: the sound arriving through the air, and the sound conducted through your skull, which reinforces the low end. A recording only has the first half. The microphone is not distorting you — you have spent your whole life hearing yourself with a bass boost nobody else ever got.

The number 48,000 comes from film

The CD’s 44,100 Hz came from fitting audio onto existing video tape. 48,000 won in professional video because it divides cleanly into the common frame rates, and that is what your browser reaches for by default.

Noise cancellation needs two microphones

A phone compares what arrives at each one. Whatever sounds the same in both came from far away and is noise; whatever arrives sooner and louder at one came from close and is your voice. That is why accidentally covering the top microphone makes a call worse even though you speak into the bottom one.

Audio decibels are negative on purpose

In digital recording 0 dB is not silence: it is the loudest value that can be represented. Everything else sits below it, in negative numbers. A healthy speaking level is around -12 dB, and reaching 0 means you are clipping.

Proximity effect is used deliberately

Directional microphones boost the low end when the source is very close. It is a physical artefact of the design, but it is exactly what produces the deep radio voice, and broadcasters exploit it by leaning in.

Questions

Frequently asked questions

Is my audio stored or uploaded?

No. Everything happens inside your browser: the sound is analysed to draw the waveform and compute the level, and nothing is sent to any server. The test recording lives in the tab’s memory and is discarded when you reload or close it.

Why does it ask for permission?

Because browsers require explicit permission for the microphone, and rightly so. The permission covers this tab only and lapses when you close it. While it is active you will see your browser’s or your system’s recording indicator.

Does it work on a phone?

Yes, on iPhone and Android, with the browser asking permission the same way. On iPhone use Safari or a recent browser; if no dialog appears, check the browser has microphone permission in system settings.

My microphone works here but not on video calls — why?

Because each application picks its own input device, separately from the system. Working on this page is a useful diagnosis: it rules out the hardware, the cable and the system permission, and leaves the problem inside that application.

What level should I be hitting?

Speaking normally, the bar should move freely without reaching the end. Barely moving means not enough gain or too much distance. Hitting the top and turning red means clipping, and that cannot be fixed afterwards by turning anything down.

What do the device values mean?

The browser reports them about your microphone. Sample rate and bits per sample define how much resolution the signal carries; latency is the delay the input adds. Automatic gain control, echo cancellation and noise suppression show as off because we deliberately ask for them off: the point is to show you the microphone raw, not what the system has decided you should hear.

What is the frequency spectrum telling me?

It splits the sound by pitch. A normal voice mostly fills the middle band, which is where intelligibility lives. A narrow fixed spike at the far left that does not go away when you stop talking is mains hum. If the right-hand side is permanently lit, there is constant hiss: a fan, or the gain set too high.

Why does my recorded voice not sound like me?

Because when you speak you hear your voice through the air and also through the bones of your skull, which reinforce the low end. A recording only captures the first part. What you hear in the recording is what everyone else hears.