Microphone test
Most microphone tests tell you the browser can hear something. This one shows you the level in decibels, draws what is arriving, records five seconds so you can hear yourself back, and lets you switch off the noise suppression your browser applies by default — because that is usually what is between you and knowing how you really sound.
Say something at your normal distance and volume.
Too loud — the signal is clipping
- Level now—
- Loudest so far—
- Background noise—
The recording stays in this page and is never uploaded.
Could you hear yourself clearly in the playback?
Reading the meter
Level now is the current loudness in dBFS — decibels relative to the loudest the format can hold, so all values are negative and zero is the ceiling.
Loudest so far is the highest peak since the test started. This is the number that matters for clipping.
Background noise is the quietest level seen, which is a fair estimate of your room and hardware noise between words. The bigger the gap between it and your voice, the clearer you sound.
| What you see | What it means | What to do |
|---|---|---|
| Level around −18 to −6 | Correct | Nothing |
| Peaks hitting 0, clipping warning | Too loud | Lower input volume, move back |
| Level below −30 | Too quiet | Raise input volume, move closer |
| Meter never moves | Nothing arriving | Wrong device, muted, or blocked |
| Level jumps up and down on its own | Automatic gain is active | Switch the processing toggle off |
| Voice clear but noisy behind it | Room or hardware noise | Run the background noise test |
The processing toggle
This is the control that makes the difference between a test and a demonstration.
With processing off, you hear the microphone as it is: the room, the hiss, the plosives. That is what you need for diagnosis.
With processing on, the browser applies the same noise suppression, echo cancellation and automatic gain a video call would. If your voice sounds fine here but bad in a call, or fine in a call but bad here, the difference is almost always this processing, and knowing which side sounds wrong tells you what to fix.
Windows 11
- Pick the right input. Settings → System → Sound → Input. The device selected here is the system default, and it is not necessarily what any given application uses.
- Input volume is in the same panel. Start at 75 and adjust from the meter above rather than by guesswork.
- Audio enhancements — Settings → System → Sound → choose your device → Audio enhancements. These sit between the microphone and everything else. Turn them off while diagnosing.
- Permission. Settings → Privacy and security → Microphone. Both “Microphone access” and “Let desktop apps access your microphone” need to be on, and browsers are listed separately.
- On Copilot+ machines, Voice Focus applies its own processing on top of everything. It is worth switching off before deciding a microphone is at fault.
macOS
- System Settings → Sound → Input. Select the device and watch the input level bar while you speak; it should move well but not sit at the far right.
- Permission. System Settings → Privacy and Security → Microphone, then confirm your browser is enabled. macOS also shows an orange dot in the menu bar whenever a microphone is live, which is a quick way to confirm something has the device.
- Another application may be holding it. macOS shares microphones, but some audio software takes exclusive control. If the browser cannot open the device, quit anything that records.
- Voice Isolation, in Control Centre during a call, changes the microphone signal system-wide for that application. If a call sounds different from this test, that is a likely reason.
If the level is right and it still sounds wrong
Three things account for most of it, in this order:
- Distance and angle. A microphone 60 cm away picks up four times less of you and just as much of the room as one at 30 cm. Speak across the microphone rather than into it and the plosives go away.
- What the microphone is resting on. A desk transmits every keystroke and mug. Anything soft under the stand removes noise no software can.
- The room itself. A hard, empty room returns your voice back into the microphone a few milliseconds late. This is the one people mistake for a hardware fault most often, and the background noise test measures its floor.
Common questions
Is my voice uploaded anywhere?
No. The audio never leaves the page. The level meter and the waveform are computed in your browser, and the five-second recording is held in memory and played back from there. There is no server to send it to, no account, and nothing is stored after you close the tab.
Why does the test switch off noise suppression by default?
Because otherwise you would be testing the browser rather than the microphone. Chrome, Edge and Safari all apply noise suppression, echo cancellation and automatic gain to microphone input by default. Those are useful in a call and misleading in a test: they can hide a fault, or create one by pumping the level up and down. The toggle lets you hear both — off is what the microphone actually sends, on is roughly what a call hears.
What level should I aim for?
Speaking normally, the meter should sit around minus 18 to minus 12 decibels, with the loudest moments reaching about minus 6. Peaking at zero means clipping, which sounds like crackling and cannot be repaired afterwards. Sitting below minus 30 means the far end will raise their volume and get your room noise along with your voice.
The test works but nobody hears me in Zoom, Teams or Discord. Why?
Those applications keep their own input selection, separate from the operating system. If you have more than one microphone — a webcam and a headset, say — the application may be listening to a different one. Check the audio settings inside the application itself, not only in Windows or macOS.
Why is there a delay before the meter moves?
The browser has to open the device and, on some systems, take it away from whatever had it. A pause of up to a second is normal. If the meter never moves at all, the input is muted at the hardware level, on a physical switch, or in the system mixer.
Can I test a Bluetooth headset here?
Yes, but expect it to sound poor. When a Bluetooth headset is used as a microphone, most systems switch it into a hands-free mode that drops audio quality sharply for everything, including what you hear. That is the protocol, not your headset, and it happens in calls too.