Mic Test: Check Your Microphone and Read the Real Input Level

By Andriy Buday · Canada / Austria / Ukraine

A microphone test answers three questions in about thirty seconds: does the browser see your mic, is sound actually reaching it, and is the level in a range that other people can hear comfortably. Everything on this page happens inside your browser tab. The audio stream opens when you press Start preview, gets analysed frame by frame, and is discarded. Nothing is recorded, nothing is uploaded, and no setting on your machine is changed.

What a browser mic test can and cannot see

The browser receives your microphone as a stream of floating point samples between -1 and +1. A level meter reads a block of those samples, measures how much energy is in the block, and prints the result in dBFS, which is short for decibels relative to full scale. Zero dBFS is the loudest value digital audio can hold, so every reading you see is negative and the ceiling sits at zero. Full scale is not a suggestion: once a sample hits it, the waveform is cut flat at the top, and the distortion that follows cannot be removed later.

From that same block of samples the meter can produce two different numbers. RMS is the root mean square, the average energy across the block, and it tracks perceived loudness closely, which is why it moves smoothly while you talk. Peak is the single loudest sample in the block, and it jumps on plosives and any other sharp sound. Our meter works from RMS because it gives a steadier reading: it computes the level 1024 samples at a time, converts it to decibels, and smooths the value so the bar does not flicker on every consonant.

What a test like this cannot tell you is absolute loudness in the room. dBFS is a digital scale, a property of the samples, not of the air. Turning it into real sound pressure in dB SPL requires a calibrated reference microphone and a known tone, which a browser page does not have. That limit does not matter for the job at hand. You want to know whether your voice arrives clean and at a usable level, and both of those are visible here.

How to run the test

The workflow is short, and the order matters because the device list only becomes useful after you grant access.

1. Open the tool and press Start preview. Your browser asks for microphone permission. Device names are hidden before that prompt, so a list that reads "Microphone 1" and "Microphone 2" simply means permission has not been granted yet.
2. Pick the input you want to test from the dropdown. Every audio input the operating system exposes to the browser appears there: built-in laptop arrays, webcam mics, USB microphones, headsets, and virtual loopback devices.
3. Speak the way you would on a call. Hold a normal conversation distance, roughly a hand's width from the microphone, and talk in your ordinary voice instead of leaning in.
4. Watch the bar and the number beside it. The fill moves with the signal, the percentage below it maps the level onto a 0 to 100 scale, and the dB value shows the same thing to one decimal place.
5. Press Stop when you are done. That closes the audio stream, disconnects the analyser, and shuts the audio context, which releases the device for other apps.

Two practical details come up often. Choosing a device in the dropdown does not change your system default, so other applications keep their own setting, and a rescan happens automatically when you plug in or remove hardware because the page listens for device change events.

Reading the level bar

The bar exists to make an abstract number readable at a glance. It maps a 60 dB window onto the full width of the meter: silence at the bottom, the point where clipping begins at the top. Anything at or below -60 dB fills nothing, -30 dB fills half the bar, and 0 dB fills it completely.

The percentage and the decibel value come from the same calculation, so they never disagree. Treat the top two rows as the danger zone: a bar that touches 100% means the loudest parts of your voice are being cut off, and every word that lands there will sound harsh on the other end.

What healthy microphone levels look like

Audio engineers settled on a few numbers decades ago, and they still describe a good speaking signal. Peaks should land between -12 and -6 dBFS during louder words. Average RMS level should sit around -30 to -18 dBFS. The gap between your voice and the noise floor should be 25 dB or more, because everything below that gap is room tone that the other side gets to hear.

Noise floor deserves its own line, since it is the part of the signal people forget. A quiet room with a decent microphone produces a floor below -60 dBFS, and a well-treated room with a good interface reaches -70 dBFS. Between -60 and -50 dBFS is acceptable. Above -50 dBFS the hiss, hum, or fan noise becomes part of your voice as far as your listeners are concerned, and no amount of gain adjustment will fix it because raising gain raises the noise with the signal.

Clipping is the mirror image of the noise floor problem. When the meter pegs at 0 dBFS, the signal is being flattened, and the result is a crackling edge that survives every later processing step. If a loud laugh pushes the bar into the top row, either back away from the microphone or lower the input gain one notch and test again.

When the meter stays flat while you talk

A flat signal is the most common reason people go looking for a mic test in the first place, and the cause is almost never a broken microphone. Work down this list, because four of the five causes take under a minute to check.

The browser is blocking the site. If you dismissed the permission prompt or chose Block in an earlier visit, the page never gets a stream at all. The padlock or shield icon in the address bar holds the per-site permission, and switching the microphone entry to Allow, followed by a reload, is usually enough to bring the meter to life.

The wrong device is selected. Docks, hubs, and headsets add inputs faster than people notice, and a laptop with a webcam, a USB interface, and a headset can easily expose four inputs where you expect one. Pick each candidate in the dropdown and speak while the meter is open; the one that moves is your real microphone.

Something is muted outside the browser. Both Windows and macOS ship a system-level switch for input devices. On Windows 11 and 10, Settings, then Privacy and security, then Microphone, holds two toggles that both need to be on: "Let apps access your microphone" and "Let desktop apps access your microphone", the second one covering desktop browsers. On macOS, System Settings, then Privacy and Security, then Microphone, lists the browser and needs a checkmark. A headset's inline mute button and a keyboard mute key produce exactly the same flat result.

Another application is holding the device. Most consumer drivers hand a microphone to the first program that opens it. A call client, a recording app, or a virtual microphone driver that is still running in the background will keep the device busy, and your browser receives a stream with nothing in it. Quitting the other program and reloading the test page clears this in one step.

The level is simply too low. If the bar twitches instead of moving, your input volume is at the bottom of its range. Raise the operating system input slider in small steps, restart the test after each change, and stop when peaks land near -12 dBFS. Keep input volume in the 60 to 80 percent range and leave the software boost off, since maxing out the slider mostly amplifies the noise floor and invites clipping.

When the test reports pure silence

There is one reading worth calling out separately, because it points somewhere specific: a meter that shows nothing at all while other inputs work. A working microphone never delivers exact digital zeros, not even in a silent room, because the noise floor of the electronics and the room always contributes something. When the samples are all zero, the signal is being stopped before it reaches the browser. That means a hardware mute switch on the headset or cable is engaged, the input volume in the operating system is at zero, a privacy toggle has revoked access, or the selected input has nothing plugged into it. Switching to another device in the dropdown is the fastest way to tell a dead input from a muted one.

Set the gain once, then stop touching it

The point of testing is to arrive at a setting you can leave alone. Pick the microphone you actually use, set input volume to a level where ordinary speech peaks near -12 dBFS, then say a sentence at each distance you normally work from and watch how far the peak moves. If leaning back drops you below -30 dBFS, you were relying on proximity rather than gain, and a call app with automatic level control will compensate by pulling the noise floor up with your voice.

Drivers and ports change things quietly. A new USB port, a system update, or a fresh headset can reset input gain, and the first symptom is a call where you sound distant. A two-minute test before an important meeting costs nothing and catches all of it, which is why the habit survives long after the troubleshooting is done.

Andriy Buday

Canada / Austria / Ukraine · SO reputation 1999 · Badges: 1🥇18🥈40🥉 · SO member since 2009