Microphone Test
Check that your microphone works and how good it sounds. See the live level in dBFS, peaks and clipping, measure the noise floor, record five seconds and play it back, and read exactly what your browser reports about the mic.
How to Use
- Press Start microphone and choose Allow when the browser asks. Until then the page shows an example made from a built-in test signal.
- Speak at your normal distance. The bar shows the level (RMS) and the white tick the peak; CLIP lights if any sample hits full scale.
- Wrong microphone? Pick another one in the Microphone list; the list fills in after you allow access.
- Press Measure noise floor and stay silent for 3 seconds to see how much hiss and background noise the mic picks up.
- Press Record 5 s to hear yourself back, and try the Raw and Call mode presets to compare the browser’s echo cancellation, noise suppression and auto gain.
Changing a box restarts the microphone with the new request; the table below shows what the browser really did.
Recordings stay in this tab’s memory only, and are deleted on Stop.
Worked Example
The example on a plain visit. The built-in test signal is one analyser window: 2,048 samples at 48,000 Hz, which is 2,048 ÷ 48,000 = 42.7 ms of sound. The squares of the samples add up to 20.4812, so the RMS is √(20.4812 ÷ 2,048) = 0.1000, and the level is 20 × log10(0.1000) = −20.00 dBFS. The largest sample is 0.2180, a peak of 20 × log10(0.2180) = −13.23 dBFS, so the crest factor is −13.23 − (−20.00) = 6.77 dB and there are 13.23 dB of headroom before clipping.
Noise floor and signal-to-noise. The quiet part of the same example has an RMS of 0.00100, which is −60.00 dBFS. The voice sits −20.00 − (−60.00) = 40.00 dB above the floor. A live 3-second measurement at 48,000 Hz uses every one of the 144,000 samples, not a few snapshots.
The common mistake: averaging decibels. If half of the quiet period reads −60 dBFS and the other half −40 dBFS (a fan switching on), the noise floor is not (−60 + −40) ÷ 2 = −50 dBFS. Energies add, not decibels: 10 × log10((10−6 + 10−4) ÷ 2) = −42.97 dBFS. The louder half dominates, which is why this tool sums the squared samples first and converts to dBFS once at the end.
Show Work
Formulas
How the Browser Hears Your Microphone, and How to Fix Access
The path of the sound. When you press Start, the page calls getUserMedia() from the W3C Media Capture and Streams specification. The browser asks your permission, opens the device through the operating system and hands the page a track. This tool feeds the track into a Web Audio AnalyserNode, which gives the latest 2,048 samples as numbers between −1 and +1 and a Blackman-windowed spectrum of them. The track is never connected to your speakers, so the test cannot howl with feedback.
Processing you did not ask for. Media Capture and Streams defines three switches that browsers apply to microphones by default for calls: echoCancellation (removes sound from your own speakers), noiseSuppression (removes steady noise such as fans) and autoGainControl (turns quiet speech up and loud speech down). They are why a mic can sound different here than in a recording program, and why the noise floor can rise between words. Not every browser lets a page turn them off; getCapabilities() says which, and the table above shows your browser’s answer.
If the browser says access is blocked. The browser and the operating system each have a switch, and both must allow it. In Chrome, open More → Settings → Privacy and security → Site settings → Microphone; the same page chooses the default microphone. In Edge, Settings and more → Settings → Privacy, search, and services → Site permissions → All sites. In Firefox, click the crossed-out microphone in the address bar and clear the block, or use Settings → Privacy & Security → Permissions → Microphone. In Safari on a Mac, Safari → Settings → Websites → Microphone. On Windows 11, Settings → Privacy & security → Microphone must have both Microphone access and Let desktop apps access your microphone on, because browsers are desktop apps. On a Mac, System Settings → Privacy & Security → Microphone must list your browser as allowed.
If no microphone is found, or the wrong one is used. A headset on a single 3.5 mm plug needs a combined headset jack. A device the system does not list (Windows Device Manager, macOS System Information) will not appear in any browser either. Laptops often default to the built-in array microphone even with a headset attached; choose the right one in the list above, or set the default in the browser’s microphone settings. If a device is unplugged while the test runs, the track ends and the page says so.
About This Tool
This microphone test shows whether sound is arriving, how loud it is in dBFS (RMS and peak), whether any sample clipped, how much noise the mic picks up in 3 seconds of quiet, and what your browser reports about the device: sample rate, channels, latency and the echo, noise and gain processing, with whether each can be switched. A 5-second recording lets you hear the result.
Privacy: the microphone is only requested when you press a button. Audio is measured live in your browser and not stored; the optional clip stays in this tab’s memory until you discard it, press Stop or leave the tab, and the microphone is released at the same moments. Nothing is uploaded.
Sources: W3C Media Capture and Streams and MediaStream Recording; W3C Web Audio API (AnalyserNode); MDN Web Docs (getUserMedia errors); AES17 (dBFS for a full-scale sine); Google Chrome Help “Use your camera and microphone in Chrome”; Microsoft Support “Windows camera, microphone, and privacy”; Mozilla Support “How to manage your camera and microphone permissions”; Apple “Customize settings for each website in Safari on Mac”.
Related tools: Speaker Test, Webcam Test, and Audio Spectrum Analyzer.
Frequently Asked Questions
How do I test my microphone?
Press Start microphone, allow access, and speak. If the meter moves with your voice, the browser is receiving sound from that microphone. Then record 5 seconds and play it back to judge how it sounds, and check that the device named in the list is the one you meant to use.
What is a good microphone level in dBFS?
dBFS counts down from the digital maximum, 0 dBFS. For speech, an RMS level somewhere around −30 to −15 dBFS with peaks staying below about −3 dBFS leaves room for laughs and loud words without clipping. A floor 40 dB below your voice (for example speech at −20 dBFS over a −60 dBFS floor) sounds clean. These are practical targets, not a standard.
Why does my microphone show no sound at all?
If the meter reads exactly −∞ the browser is receiving digital silence: the mic is muted by a headset button or switch, muted in the system sound settings, or blocked by the system privacy setting. If the level is just very low (below −70 dBFS), you probably have the wrong device selected or the input gain turned down.
Should echo cancellation, noise suppression and auto gain be on?
For calls through speakers, yes: they stop the other person hearing themselves and keep your level even. For music, singing or measuring, turn them off, because noise suppression can cut quiet sounds and auto gain raises the noise floor between words. The table shows whether your browser lets this page switch each one.
Is my voice recorded or uploaded?
No. The meter, spectrum and noise measurement run in your browser on the live stream and keep nothing. The optional 5-second clip lives only in this tab’s memory and is deleted when you press Discard or Stop, or leave the tab. Nothing is sent to this site or anyone else.
Common Use Cases
Before a video call
Confirm the headset, not the webcam’s built-in mic, is selected, and that speech reaches about −20 dBFS without lighting CLIP.
Find a muted or dead mic
Exactly −∞ dBFS for 2 seconds means digital silence: a mute switch or a system block, not a broken capsule.
Compare two microphones
Measure the noise floor of each: −65 dBFS against −45 dBFS is a 20 dB difference, ten times the noise amplitude.
Set a podcast or stream level
Raise the input gain until your loudest laugh peaks around −6 dBFS and the clip counter stays at 0.
Check what processing a browser adds
Switch to the Raw preset and watch the noise floor: auto gain lifting it from −62 to −48 dBFS makes the hiss 5 times larger.
Last updated: