Microphone test

A microphone test that stops at the meter is only half an answer

Almost every microphone test online answers one question: is any signal arriving? That question is worth answering, and it is the one people ask when a call is about to start and nobody can hear them. But it is a low bar. A microphone can be plugged in, selected, unmuted, and producing a lively meter while the audio it captures is unusable — clipped into crackling, buried under a fan, hollow because the laptop lid is between you and the built-in capsule, or picking up the room instead of your voice because the headset is pointed the wrong way.

None of that shows up in a bar that moves. All of it is obvious within two seconds of hearing yourself played back. That is why this page records a short take and plays it to you rather than stopping at the meter: the meter tells you the signal is arriving, the playback tells you what arrived.

Raw microphone, or the way a call hears you

There is a detail here that quietly ruins most browser-based microphone tests. When a page asks for microphone access, browsers apply a chain of processing by default: echo cancellation, noise suppression, and automatic gain control. Automatic gain is the important one. It continuously raises quiet input and lowers loud input, which means a level meter placed after it shows you the browser's correction rather than your microphone's actual output. Speak quietly and the meter rises to meet you. Speak loudly and it settles back down. The result looks reassuring and measures nothing.

This page therefore opens the microphone raw by default, with all three off. What the meter shows is what the device produced. The processing can be switched back on with one control, and doing so is genuinely useful for a different question — video-call software applies similar processing, so turning it on tells you roughly how a meeting will hear you. Two settings, two real questions, and the page says which one you are looking at.

What the decibel numbers mean

The figures are given in dBFS, decibels relative to digital full scale. Zero is the loudest value a sample can hold, and every real level is a negative number. Peaks landing between roughly −24 and −3 dBFS sit where recording software expects them, with room left over for a sudden louder word. Above −3 there is almost no room left. At 0 the waveform is being flattened against the ceiling, which is what clipping is, and no amount of later editing restores what was cut off.

Peak and average are both shown because they answer different things. Peak catches the single loudest instant, which is what clips. The average, measured as RMS, describes how loud the recording feels overall. A voice with a healthy average and occasional peaks near the top is normal; an average close to the peak usually means something is compressing the sound before it reaches the page. The verdict in words is a convenience, and the numbers are printed next to it on purpose: the bands are a convention, and a convention presented without its measurement starts to look like a fact.

When nothing reaches the page

The most common cause is not a broken microphone but the wrong one. Browsers frequently choose a built-in laptop microphone over the headset you just connected, so the device name shown on this page is worth reading before anything else. After that, check the operating system's sound settings for a muted or zero-level input, and look for a physical mute switch on the microphone, the cable, or the headset earcup. If a video-call application is still running in the background it may be holding the device exclusively, in which case the browser can see the microphone but cannot open it — this page says so specifically rather than reporting a generic failure.

The audio never leaves this tab

Your browser opens the microphone, this page measures the samples, and they are discarded when you navigate away. There is no upload, no account, and no server involved at any point. The downloadable file is a WAV assembled in this tab from the exact samples your microphone produced — which is also why the playback above and the download are the same bytes, rather than a preview of something you have not heard. If you would rather confirm that than take it on faith, the guide to verifying local processing explains what to watch for in your own browser's network panel.

When the recording is the point, not the microphone

This page exists to judge equipment, and it throws the audio away by design. When the levels look right and you want to keep what you say next, the browser audio recorder is the sibling that writes a real file instead of a short sample, with pause and resume and no length limit. Once you have a recording you want to change rather than keep, the work moves to tools built for files. To cut an exact section out of a take, the audio trimmer works from the waveform and writes a new file. Long pauses at the start and end of a recording are worth removing before anything else, and the silence remover finds them by threshold rather than by eye. If the file needs to arrive somewhere as a different format, the audio converter handles that locally. And when a recording has to be shown rather than heard — in a post, a thumbnail, a podcast page — the waveform generator draws it as an image.