Watch your voice arrive as a waveform, check the level, then record a few seconds and hear yourself back.
Microphone off
Nothing is captured until you start it, and nothing is ever sent anywhere.
Loudest peak—
Clipped frames0
Speak normally. Peaks between -12 and -6 dBFS are the target.
—dBFS noise floor
Stay quiet for a few seconds and the floor settles. Below -60 dBFS nobody will hear your room.
Say a sentence at your normal distance, then play it back. Hearing yourself is the only reliable way to judge hiss, echo and room noise.
0 dBFS is the loudest a digital signal can go, so every reading here is a negative number counting down from that ceiling. Speaking normally, you want peaks landing between -12 and -6 dBFS, which leaves headroom for a laugh without touching the limit.
Browsers apply echo cancelling, noise removal and automatic gain by default. That helps on calls and hides real faults. Switch to raw input to hear the microphone itself, including the hum, hiss or crackle that processing was papering over.
Moving twice as far from a microphone costs you roughly 6 dB of signal while the room noise stays where it was. A cheap microphone a hand's width away will out-perform an expensive one across the desk, every time.
Give the page permission once, watch the meter while you talk, then record a sample so you hear what everyone else hears.
Press Start microphone. Your browser asks for permission, and the recording indicator in the tab or address bar confirms it is live.
If you have more than one input, pick the right one from the device list. The list only fills in with real names after permission is granted.
Speak normally at your usual distance. The waveform should move clearly and the peak reading should land somewhere around -12 dBFS to -6 dBFS.
Flat line means nothing is arriving. A meter pinned at 0 dB means you are clipping, which is the distortion people hear as crackling on calls.
Stop talking for a moment to let the noise floor settle, then record a few seconds and play it back. This is the only reliable way to judge hiss, echo and room noise.
If the playback sounds distant or hollow, move closer, or check you are not recording from a laptop mic while wearing a headset.
What the numbers mean, and what to try when the meter stays flat.
Browsers will not open a microphone without an explicit prompt, and that prompt is the only way this page can read audio at all. Permission applies to this tab only and ends when you close it or press stop.
If you dismissed the prompt by accident, click the padlock or camera icon in the address bar and reset the microphone permission for this site, then reload.
Work through these in order:
Speaking normally, peaks around -12 dBFS to -6 dBFS give you clear speech with room left for a laugh or a raised voice. Average level sitting near -18 dBFS is healthy.
If peaks hit 0 dB the signal is clipping and permanently distorted, so turn the input gain down. If you cannot get above about -30 dBFS, turn gain up or move closer.
The noise floor is the level of your room and equipment when nobody is speaking: fans, traffic, air conditioning, and hiss from the microphone itself.
Below about -60 dBFS is quiet enough that nobody will notice. Around -50 dBFS is audible on a call. Above -40 dBFS people will hear your room before they hear you, and no amount of software noise removal fixes that cleanly.
Yes. Anything your operating system lists as an input appears in the device list, including USB condenser mics, headset booms, webcam mics, audio interfaces and Bluetooth earbuds.
Choose the device before you start, or switch it mid-test and the page will restart the stream on the new input.
Most Bluetooth headsets cannot send and receive high quality audio at the same time. The moment the microphone opens, the connection drops to a narrow voice profile, music quality collapses and everything sounds thin and muffled.
That is a limitation of the Bluetooth profile rather than a fault. A USB or wired headset avoids it entirely, which is why it is the usual recommendation for calls and recording.
Automatic gain control is the usual cause. Browsers apply it by default to keep call levels even, and it fights any gain you set yourself.
Distance is the other cause, and it matters more than people expect: doubling your distance from the mic cuts the level by roughly 6 dB. A microphone a hand's width from your mouth will always beat one across the desk.
No. The waveform, the meters and the recording all stay in your browser's memory. Nothing is written to a server, nothing is stored on disk unless you save it yourself, and reloading the page discards it.
Chrome, Edge, Firefox, Safari and Opera all support microphone access on desktop, and Chrome and Safari support it on Android and iOS.
Access requires a secure connection, so a page served over plain HTTP cannot open your microphone at all. That is a browser rule rather than a site setting.
It tests everything up to the moment your voice leaves the browser: the device, the driver, the level and the sound quality. If the waveform moves and the playback sounds right, your microphone is working.
If it works here but not in a specific app, the fault is that app's device selection or permissions rather than your hardware, which is a much quicker thing to fix.
Every tool runs in this tab, reads your hardware directly through the browser, and stores nothing.
Check buttons, analog sticks, triggers, the D-pad and rumble, and measure stick drift.
SpeakersLeft and right channel checks, a tone generator, and a 20 Hz to 20 kHz hearing sweep.
KeyboardLight up a full layout key by key, read live key codes, and test N-key rollover.
MouseButtons, scroll wheel, movement trace, polling rate in Hz, and a double-click chatter check.
WebcamLive preview, real resolution and measured frame rate, camera switching and snapshots.
ScreenDead pixel hunt, backlight bleed check, gradient banding, and a measured refresh rate.