Microphone Test

Press start, say something, and this shows you the waveform, the level, and the technical details of what your browser has actually opened. It tells you plainly whether the microphone is working, and if the input is overloading it says that too.

How to use it

1. Press Start and allow microphone access. Nothing is recorded at any point.
2. Speak at a normal volume and distance. The waveform should show clear movement and the meter should sit around the middle.
3. If the clipping warning appears, turn down the input gain in your system sound settings — the signal is being cut off.
4. Switch devices with the selector to confirm which microphone the browser is really using.
5. Hear yourself unmutes a monitor path. Use headphones, or you will get feedback.

The latency number nobody can give you honestly

“What is my microphone latency” is the most common real question behind a mic test, and it is the one where most tools either stay silent or invent a figure. Here is why it is hard.

Latency is the time between a sound existing in the room and you hearing it come back out. That journey has at least six stages: the microphone and its preamplifier, the analogue-to-digital converter, the operating system’s capture buffer, the browser’s own processing buffer, the operating system’s output mixer, and finally the digital-to-analogue conversion and the speaker.

A browser can see exactly two of those. AudioContext.baseLatency reports its own processing buffer, typically 128 to 512 samples, which is around 3 to 11 milliseconds at 48 kHz. AudioContext.outputLatency reports the platform’s estimate of the path from the browser to the speaker, and it is unavailable in several browsers entirely. Everything else — the capture side, the OS mixer, the hardware — is invisible.

So the figure shown here is honest but partial, and it is labelled as such. In practice a wired setup on a decent machine lands somewhere around 20 to 40 milliseconds of real round-trip latency, of which the browser can only account for perhaps a third. Bluetooth is a different world: the codec alone adds 100 to 300 milliseconds, which is why musicians do not use Bluetooth headphones for monitoring and why a video call over AirPods feels subtly out of step.

If you need to know your true round trip, the only reliable method is a loopback measurement: play a click, record it through the microphone, and measure the gap in the resulting waveform. That requires recording, which this page deliberately does not do.

Why the waveform is drawn the way it is

A detail worth mentioning because it changes what you see. The analyser hands over 2048 samples per frame, and the canvas is perhaps 400 pixels wide. That is five samples per pixel column.

The naive approach plots one sample per column and discards the rest, which loses the peaks — a transient that lasts three samples vanishes entirely, and a loud clipped passage can look calm. This tool keeps the minimum and maximum of every column and draws a vertical line between them. The result is the familiar filled waveform shape, and more importantly it means a spike is always visible. That is the same technique every audio editor uses, and it is why the clipping detector and the visual agree with each other.

What it does not do

It does not record, so there is nothing to play back beyond the live monitor. It cannot measure true round-trip latency, as above. It cannot fix a driver problem or change your system gain — it can only show you that the gain is wrong. Device labels are blank until permission is granted, which is a privacy protection in the browser rather than a fault. And switching tabs stops the stream deliberately, to release the microphone for other applications.

Nothing is uploaded. No audio is written to disk or transmitted anywhere.

What a healthy reading looks like

Measurement	Expected
Silent room level	Under 1%
Normal speech level	10–40%
Peak on loud speech	Under 90%
Clipping warning	Should never appear
Sample rate	44,100 or 48,000 Hz
Channels	Mono for most headsets

Common causes of a dead microphone

Symptom	Usual cause
No permission prompt at all	Site permission already blocked in browser settings
Prompt appears, then silence	Wrong device selected, or muted in the OS
Works in one app only	Another app holds exclusive access
Very quiet even when shouting	Input gain set near zero, or a broken cable
Clipping at normal volume	Input gain or boost set too high
Crackling and dropouts	USB power saving, or a failing cable