Microphone test

What this microphone test measures

Almost every microphone test works the same way: a bar wobbles when you speak, and you are told the mic works. That answers the least interesting question. Your microphone was probably never broken — it was too quiet, or clipping, or picking up the air conditioning, or the browser was listening to the wrong input. A moving bar cannot tell you which.

So this page gives you numbers instead of reassurance, without sending anything anywhere: it opens the stream with getUserMedia, reads the sample values in a Web Audio analyser a few dozen times a second, and draws them. There is no upload endpoint in the code and no server to receive one.

• Peak and average level in dBFS — the actual numbers, not a coloured bar, with a plain verdict: too quiet, good, very hot or clipping.
• Your noise floor — the quietest half-second out of the last three, the honest measure of what your room adds.
• A device picker, so you can confirm you are testing the headset and not the laptop lid, plus sample rate, channels, per-channel levels for a stereo interface, and whether echo cancellation and automatic gain are on.
• Recording with playback — up to sixty seconds, played in the page, saved to disk only if you ask.

How to test a mic, step by step

About thirty seconds, start to finish. If you landed here mid-panic before a call, this is how to test a mic properly:

• Press Allow microphone. Nothing runs before you do — no page should open your microphone unasked.
• Check the device picker shows the input you meant to use. Browsers hide labels until permission is granted, which is why the list fills in only now.
• Say nothing for three seconds and watch the noise floor settle. That number is your room.
• Speak normally at the distance you would actually use, not leaning in to test it, and aim for peaks between -12 and -6 dBFS.
• Record a few seconds and listen back, reading something with hard consonants. The p and b sounds expose popping that a meter never shows.

Microphone test playback: the part most tools skip

A meter tells you the signal exists at a sensible level. It cannot tell you that your headset is compressing everything into a tinny mid-range, or that every plosive lands like a thump. Only listening does, which is why microphone test playback is the step worth insisting on rather than skipping.

Listen for four things: sibilance, where s-sounds turn harsh; plosives, where p and b thump; room, the faint echo of hard walls; and background noise. Each has a different fix, and you cannot choose until you have heard which it is. Play the clip through something other than the headset you recorded on, too — cheap headsets flatter their own recordings.

Reading dBFS, and what a mic quality test really measures

dBFS means decibels relative to full scale. Zero is the loudest value a digital signal can hold, so every real level is negative, and the scale is logarithmic: -6 dBFS is half the amplitude of 0, -12 a quarter, -20 a tenth. It is the only honest form a mic quality test can take in a browser, because it describes the signal in the pipeline rather than the sound in your room.

Why -12 to -6 dBFS is the target

Headroom. Speech is spiky: the average level of a person talking sits fifteen to twenty decibels below their peaks, so a recording peaking at -6 is comfortably loud with room left for the one sentence you say with emphasis. Push peaks to -1 and that sentence squares off and distorts, and distortion is the one problem you cannot fix afterwards — the information is gone, not hidden. Quiet is the opposite: recoverable, but every 6 dB you add back in editing raises the hiss and the room by 6 dB too.

Your noise floor is the number nobody checks

The noise floor is everything that is not you: the mic’s own electronics, the fan, the aircon, the street. This page finds the quietest half-second in the last three, so a pause between sentences is enough to read it.

Below -60 dBFS is a genuinely quiet room; -55 to -50 is normal and perfectly usable. Above -50 something is running, and above -40 it will be audible in every recording you make, with gain no help at all, because raising your voice raises the noise with it. What matters is the gap rather than the number — forty decibels between your speech and the floor sounds clean, fifteen sounds like a call from a server room. The usual culprits are a laptop fan that spins up during video calls, an air conditioner, and an open window.

Too quiet is nearly always input gain, not a broken mic

If the meter barely moves while you are clearly speaking, the microphone is almost certainly fine. Input gain is a per-device operating-system setting, and it gets moved by accident — by an app that adjusts it, a driver update, or an automatic-gain algorithm deciding your quiet room needed less amplification. On Windows: Settings, System, Sound, your input device, raise Input volume. On macOS: System Settings, Sound, Input.

Distance matters as much as the slider: halving it, 40 cm to 20 cm, gains you roughly 6 dB without amplifying the room with you. If the level is still low at maximum, suspect hardware — a mute switch on the cable, or a boom mic rotated away from your mouth.

How to test a microphone on Windows 11

Windows has its own check: Settings, System, Sound, pick the input, speak, watch the Test your microphone bar. It confirms Windows hears you, but gives no dBFS level, no noise floor, and no sign of what a given application will receive — which is where nearly all real Windows 11 microphone problems live.

The per-app setting that breaks one app and not another

If your mic works in one program and not another, this is almost always why. Go to Settings, Privacy & security, Microphone. Check that Microphone access is on, then that Let apps access your microphone is on, then scroll down: there is a separate list of individual apps, and below it a further switch called Let desktop apps access your microphone. Browsers sit in the app list, installed desktop software under that last switch, and turning off one while leaving the other on produces exactly the symptom people describe as “my mic works in Zoom but not in the browser”. The macOS equivalent is System Settings, Privacy & Security, Microphone, where the browser counts as one application — and Safari keeps a second, per-site permission under Settings, Websites, Microphone.

Why a mic test online beats the built-in meter

A mic test online has one real advantage over the operating system’s own panel: it tests the path your browser actually uses. If you are about to join a call in Chrome, what matters is what Chrome receives after the OS, the driver and the browser have each had a turn, and a settings panel measures earlier in that chain. That is the argument for a mic test tool in a browser tab — not convenience, but that it measures the same signal your call will. The argument against most of them is that they upload your recording in order to play it back, leaving a clip of your voice on someone else’s disk under retention rules you will never read. Browsers have replayed audio locally for a decade.

What this tool deliberately does not do

It cannot tell you how loud you are in dB SPL, the measurement a sound level meter gives. dBFS describes the signal after your microphone, preamp and gain have altered it, so two mics in the same room at the same real loudness read completely different values. Converting to SPL needs a calibrated reference, which no web page has on its own; the decibel meter gives an estimate, says so, and lets you calibrate it against a real meter.

It cannot see a hardware mute either. A switch on a headset cable, a mute button on a USB mic or a muted channel on a mixer all produce exactly what a silent room produces, so if the meter is flat while you are speaking, check for a physical switch first.

And it cannot tell you what processing already happened. The browser reports whether it asked for echo cancellation, noise suppression and automatic gain, shown under the meter — but operating systems and drivers apply their own first and do not announce it. An implausibly clean noise floor may be real, or a noise gate you cannot see. There is no webcam mode either, on purpose, and no clip is kept: reload and it is gone. If you do need to keep a section of something already recorded, the audio trimmer cuts it in the browser too.

Where VoiceSnap Pro fits

This page exists because we build a dictation app and kept needing it. VoiceSnap Pro is a voice-to-text app for macOS and Windows: hold one keyboard shortcut, speak, and clean punctuated text appears in whatever field your cursor is already in. It punctuates and paragraphs as it transcribes, strips filler words, and saves every dictation to a searchable notes library. It is a one-time $39 purchase, not a subscription.

The honest connection: a dictation app is only as good as the audio you hand it. Many of the transcription errors people blame on software are a level or noise-floor problem — a mic at -35 dBFS with a fan running gives any recogniser a much harder job than the same voice at -9 in a quiet room. Fix that here first. VoiceSnap Pro has not shipped yet; there is no download, only a list. Join the waitlist for one email on release day, or try browser voice typing meanwhile to see how dictated text reads.