A meter that reads in dBFS
RMS level, the instantaneous peak, and a peak-hold marker on a 60 dB scale, read from the Web Audio analyser on every frame rather than smoothed into a vague bouncing bar.
Find out whether your microphone works before the call starts. Speak and watch a live level meter, catch clipping, switch between inputs, turn the browser's voice processing on and off, and record a few seconds to hear what other people will hear.
Nothing is requested until you press Start.
Off. Nothing has been requested yet.
Browsers hide device names until a site is allowed to use the microphone. The list fills in after you press Start.
What the page asks the browser for. Each change reopens the microphone, so record a clip with a step on and off to hear the difference.
Removes your own speakers from what the mic hears. Essential on speakers, but it can thin out music.
Browser reports: after Start
Filters steady background sound such as fans and hum. Pushed hard, voices can sound watery.
Browser reports: after Start
Turns the level up and down for you. It rescues a quiet mic and lifts room noise between words.
Browser reports: after Start
Audio is measured inside this tab and never uploaded. The live microphone is never played through your speakers, so it cannot feed back into a howl.
Press Start mic test and allow access when the browser asks.
Only for this site, and only when you press Start. Nothing is requested when the page loads.
The bar, the peak marker, and the verdict follow your voice as you talk.
A short clip plays back from memory. The live microphone never goes to your speakers.
Level, peak, and clipping are measured from the stream itself. Sample rate and channels are what the browser reports for the open track.
The last ten seconds of level are drawn here once the microphone is on, so you can see how steady your voice is from word to word.
Loudest peak, last 3 s: –
The meter shows level; only your ears can judge the sound. Record a few seconds, then play it through the speakers or headphones you will use on the call.
Start the microphone first. The recording is made from the same input the meter is reading.
Held in this tab's memory only: recording again, reloading, or leaving discards it, and download is the only way to keep it. Playing it through speakers will move the meter too; headphones avoid that.
The meter shows how strong the signal is and whether it clips. Whether it sounds good is for your ears to judge, which is what the recording is for.
Readings are relative to digital full scale, where 0 dBFS is the largest value the system can hold. They are not dB SPL, and they move with the input level set in your operating system.
Echo cancellation, noise suppression, and automatic gain control run before this page sees anything. The readout shows what the browser says it applied, which is not always what was asked for.
A clip lives only in this tab. It is replaced when you record again and gone when you leave or reload. Nothing is uploaded; download is the only way to keep it.
2,048-sample analyser window · meter floor -60 dBFS · clipping at 99% of full scale · verdict over the last 3 s · nothing uploaded
With the microphone confirmed, check that your script actually fits the slot before you press record.
Sound a steady reference note to tune to before you sing or play into the microphone.
The other half of a call check: confirm the camera, its resolution, and the frame rate it delivers.
When you press Start, the page opens your microphone with getUserMedia and gives each of its channels a Web Audio analyser, which hands over the latest 2,048 samples on every animation frame. The root mean square of that slice is the level, the largest sample is the peak, and new samples at or above 99 percent of full scale are counted as clipped. On a stereo input the louder channel sets the level, so a mic plugged into one side of an audio interface reads true. Readings are shown in dBFS on a 60 dB scale, and the verdict is drawn from the loudest moments of the last three seconds, so it describes how you sound rather than one stray click. The analysers are never connected to your speakers.
The browser asks once, for this site only. Nothing is requested when the page loads. If you have more than one input, pick the one you plan to use from the list that appears after you allow access.
Talk at your normal distance and volume. The bar shows the level in dBFS, the marker holds the latest peak, and the verdict says in plain words whether the signal is missing, too quiet, healthy, running hot, or clipping.
Record three, five, or ten seconds and play it through the headphones or speakers you will use. Then switch noise suppression or automatic gain control off and record again to hear what the browser was doing to your voice.
RMS level, the instantaneous peak, and a peak-hold marker on a 60 dB scale, read from the Web Audio analyser on every frame rather than smoothed into a vague bouncing bar.
No signal, too quiet, good level, running hot, or clipping, judged on the loudest moments of a rolling three-second window, so the pause between two words does not flip the answer.
Samples at or above 99 percent of full scale are counted as the meter reads them, on every channel, so a crackle someone mentioned on a call can be traced to a gain setting instead of being blamed on the connection.
Choose any connected microphone, headset, webcam mic, or audio interface. The list refreshes when you plug something in, and the page shows which device the browser actually opened.
Request echo cancellation, noise suppression, and automatic gain control on or off, and see what the browser reports it applied. These steps change your sound more than most people expect.
The live microphone is never sent to your speakers, which would howl. You hear yourself through a short recording instead, held in memory and downloadable if you want to keep it.
Work through four layers in order. The browser needs permission for this site. The operating system needs to allow the browser to use the microphone. The right input has to be selected, since a laptop with a headset and a webcam attached often has three. And no other app can be holding it, because some call and recording apps take the microphone exclusively. If this page shows an error, its wording names the layer that refused. If the meter opens but stays flat, look for a mute switch on the headset cable or a mute key on the keyboard.
In Chrome and Edge, click the site controls icon at the left of the address bar, set Microphone to Allow, and reload. In Firefox, click the permissions icon in the address bar, clear the blocked microphone setting, and try again. In Safari on a Mac, open the Safari menu, choose Settings for This Website, and set Microphone to Allow. Then check the operating system: on macOS, System Settings, Privacy & Security, Microphone must show your browser switched on; on Windows, Settings, Privacy & security, Microphone needs microphone access and desktop app access both turned on.
To the operating system, a call app and a browser are separate applications, and each needs its own permission, so macOS or Windows can allow Zoom while still blocking Chrome. The browser then keeps a per-site permission on top of that. A call app that is still running may also be holding the microphone or may have changed the default input, so quit it and try here again. Permission given to a site in one browser does not carry over to another browser either.
Usually the input level in the operating system is low: on Windows it is under Sound settings, the input device's properties, and on a Mac under System Settings, Sound, Input. Distance matters more than people expect, since every doubling of the distance to the microphone costs about 6 dB. A headset boom swung away from the mouth, or a laptop mic under a half-closed lid, also reads low. Automatic gain control normally hides a quiet input, so with it switched off here the meter shows the raw level your hardware delivers.
dBFS is decibels relative to full scale. 0 dBFS is the largest value digital audio can hold, so every real reading is negative. For speech, aim for peaks around -12 to -6 dBFS, with the average level well below that. That leaves headroom for a laugh or a hard consonant without touching the ceiling. Peaks reaching 0 dBFS are never good. dBFS says nothing about how loud you are in the room, because the input volume in your operating system shifts every reading up or down.
Clipping happens when the signal asks for more than the largest digital value, so the tops of the waveform are cut flat. It sounds like crackling or harsh distortion on loud syllables, and it cannot be repaired afterwards. This page counts samples at or above 99 percent of full scale as the meter reads them. To fix it, lower the input gain in the operating system or on the audio interface, move back from the microphone, or turn automatic gain control on so the browser rides the level for you.
Nothing is uploaded. The meter reads the audio inside this tab through the Web Audio API and discards each slice once it has been measured. A recording exists only after you press Record, lives in this tab's memory, and is thrown away when you record again, leave, or reload. Download is the only way to keep it, and that saves to your own device. Pressing Stop releases the microphone, and the browser's recording indicator should go out.
They are the browser's voice processing, applied before any page or call app receives the audio. Echo cancellation removes the sound of your own speakers from the microphone signal, which is what stops other people hearing themselves. Noise suppression filters steady background sound such as fans and hum. Automatic gain control raises and lowers the level as you speak. All three help on calls, but they can thin out music or make a voice sound watery, which is why it is worth recording a clip with each one switched off.
When a Bluetooth headset's microphone is in use, most headsets leave the high-quality music profile and switch to the hands-free profile, which carries two-way audio in a narrow band, often sampled at 8 or 16 kHz. Your voice sounds muffled and what you hear in your ears drops in quality at the same moment. The sample rate shown here may still read 48 kHz, because the system resamples before the browser sees it. For a call that matters, use the laptop or a USB microphone for input and keep the headset for listening.
It is a good first check: if the meter moves and the playback sounds right, the hardware, its driver, and the operating system are all working. It cannot prove the call app is set up the same way, because permission is granted per site and per app, and each app keeps its own choice of input device and its own processing. A call that runs in the browser, such as Google Meet or Teams on the web, needs the same site permission this page did. A desktop app needs its own permission and its own input selection.
Explore the growing collection for calculations, documents, writing, and everyday work.