Mic Test
See your microphone working: a live level meter, the waveform of what is arriving, and a five second recording you can play straight back. The reading is in dBFS, taken from the raw samples the browser hands over, so it tells you the two things a broken microphone actually fails on: whether anything is arriving at all, and whether it is loud enough to record. Noise suppression, automatic gain and echo cancellation all start off, because a browser that quietly tidies the sound up first would flatter a bad input, and echo cancellation is a switch you can turn back on to hear what a call will sound like. Nothing is uploaded, the recording lives in memory only, and the stream closes the moment you press Stop or leave the page.
Test your microphone
Your browser asks for permission first. Nothing is recorded until you press Record, and nothing is uploaded at any point.
Nothing is uploadedThe recording stays in memoryThe stream closes when you stopNo account, nothing installed
The microphone is closed. Nothing is being listened to until you press Start.
- Record five seconds and it plays back here. It stays in memory and is replaced by the next recording.
- no readings yet: press Start
- no input open yet
- press Start and speak to find out
Why the browser asks permission first. A page cannot listen until you allow it, and the prompt is your browser's, not this site's. Allow it, speak, and the meter should move within a second. If it does not, the Device row above names the input the browser actually opened, and picking another from the list is the usual fix. Echo cancellation, noise suppression and automatic gain all start off, because a test that lets the browser tidy the sound up first would flatter a bad microphone. Switch echo cancellation on to hear what a call will sound like instead.
What the numbers mean
Every reading is in dBFS: decibels relative to full scale, where a sample at the maximum the format can hold is 0 dB and everything quieter is negative. It is a measure of the recording, not of the room, so it cannot be compared with the dB figure on a sound level meter. The verdict needs one more thing besides the level: a reading that never changes is background noise whatever its height, so the tool calls that No signal too, and says which of the two it found.
| Peak reading | What it means | What this tool says |
|---|---|---|
| -1 dB and above | Against the ceiling. 0 dBFS is full scale, the loudest a sample can be, so a louder word than the one you just said will distort. | Working, with a warning |
| -12 dB to -6 dB | A healthy recording level with room to spare above it. | Working |
| Below -30 dB | Audible but thin, and the room comes through with it. | Very quiet |
| At or below -60 dB | Nothing is arriving from this input. | No signal |
Accuracy. Levels are measured from the raw samples the browser provides, in decibels relative to full scale. Playback is the same audio returned from memory. No audio is uploaded, and the stream closes when you stop or leave the page. There is no sound pressure reading here and there cannot be: that needs a calibrated microphone with a known sensitivity, and a browser will not say what your microphone or your operating system did to the signal before handing it over.
Common questions
- Why does the browser ask permission before the mic test works?
- Because no web page is allowed to listen without being told it may. The prompt comes from your browser, not from this site, and the choice is stored by your browser against this address. Allow it, speak, and the meter should move within a second. If you blocked it by accident, click the microphone or lock icon in the address bar, set the microphone back to allow, and press Start again. Until you allow it, the browser will not name your inputs at all, so the list offers only System default. After the first Start it fills with their real names, and you can switch between them without asking permission again.
- The meter does not move when I speak. What now?
- Check the Device row under the tool first: it names the input the browser actually opened, which is often not the one you expected, because the system default can be a webcam or a monitor rather than your headset. Pick a different input from the list and the stream reopens on it. If the meter still sits flat, the verdict tells you which of the two failures you have. No signal means nothing is arriving at all, so look for a mute switch on the device or the cable and check the input is not muted in your system sound settings. A steady reading that never changes means something is arriving but it is not your voice: that is usually the wrong input, picking up a fan.
- Is my audio uploaded or stored anywhere?
- No. Every reading is calculated in the page from samples the browser hands to it, and the five second recording is held as a blob in memory and played back from there. It is not sent anywhere, it is not written to disk by this page, and it is thrown away when you record again, press Discard, or leave the page. The only thing this tool remembers between visits is whether you had echo cancellation on or off, stored in your own browser. Stopping releases the microphone track, which is what turns the recording indicator off.
- Should echo cancellation be on or off?
- On for calls, off for recording, and the switch reopens the stream so you can hear the difference. Echo cancellation exists to stop the far end of a call hearing itself come back through your speakers, and it does that by subtracting what your speakers are playing from what your microphone hears. That processing also thins the sound and can duck your voice when it guesses wrong, which is why it belongs off when you are recording rather than talking. Noise suppression and automatic gain are requested off in both positions here, so the level you see is the level the microphone produced.
- Why does my Bluetooth headset sound worse once the mic is on?
- Because a Bluetooth headset cannot send and receive high quality audio at the same time. While it is only playing, it uses a music profile with plenty of bandwidth. The moment something opens its microphone, it switches to a hands free profile that splits the bandwidth in both directions, and the music side drops to roughly telephone quality. That is the headset and the protocol, not your computer or this page. A wired headset or a USB microphone avoids it entirely, and on many systems you can pick the built in microphone while keeping the headset for playback.
- What is dBFS, and what is a good level?
- dBFS is decibels relative to full scale: a sample at the maximum the format can hold is 0 dB, and everything quieter is a negative number. Speech peaking somewhere between -12 and -6 dBFS is a healthy recording level, loud enough to sit above the room and with room to spare before it distorts. Below -30 dBFS it is thin and the room comes with it. At or below -60 dBFS nothing is really arriving. This is not the same as the dB figure on a sound level meter, which measures pressure in the air and needs a calibrated microphone with a known sensitivity, so no browser can report it.
- What is the noise floor, and why does it matter?
- The noise floor is where your room and your microphone sit between words, and this tool reads it as the average of the quietest tenth of the last twenty seconds of readings. What matters is the gap between it and your voice. A peak of -8 dBFS over a floor of -58 dBFS is fifty decibels of headroom and will record cleanly. The same peak over a floor of -16 dBFS means a fan, a laptop, or a hissy preamp is nearly as loud as you are, and no amount of editing later will separate them. The tool says so when the gap gets tight.
- Does this work on a phone, and does it need an app?
- It works in any current mobile or desktop browser that supports getUserMedia, and there is nothing to install. Two things need to be true: the page has to be on a secure connection, because browsers only open a microphone over https or on localhost, and the browser itself has to be allowed to use the microphone at the operating system level. On iOS and Android that is a permission on the browser app, checked in the system settings, separate from the permission you grant this page.
Levels are measured from the raw samples the browser provides, in decibels relative to full scale. Playback is the same audio returned from memory. No audio is uploaded, and the stream closes when you stop or leave the page.