See your microphone level live, then record 5 seconds and play it back. Nothing is uploaded.
To test your microphone online, press Start, allow it in the browser prompt and say a few words: if the level bar moves and the verdict says Level is OK, the microphone works and is loud enough. For proof by ear, press Record 5 seconds and play back: you hear exactly what a recording app would get. The page asks for the microphone only when you press Start, never when it loads. Pressing Stop, or leaving the page, switches the microphone off. If the bar stays flat, the verdict names the likely cause: no signal means nothing arrives at all (muted, unplugged or the wrong device chosen), very quiet means a signal arrives but speech is faint.
About 14 times a second the page reads the latest 4096 audio samples, each a number from -1 to 1. The level bar shows their root mean square (the average loudness) in dBFS, decibels relative to full scale, where 0 is the loudest value a sample can have. The formula is 20 x log10(level). A sample level of 0.5 is -6.0 dBFS, 0.1 is -20 dBFS and 0.01 is -40 dBFS. The bar is empty at -60 dBFS and full at 0. Peak hold keeps the highest peak for 1.5 seconds before it follows the signal down. The verdict uses the loudest values since you pressed Start. No signal: the loudest peak stayed below -70 dBFS. Very quiet: the loudest average level stayed below -45 dBFS. Clipping: peaks at -1 dBFS or higher in 2 or more readings. Otherwise: Level is OK. These cut-offs are this tool's own rule of thumb, not a standard, and a short click between two readings can be missed.
Both switches ask the browser to apply its own processing before the audio reaches this page. Echo cancellation removes speaker sound that the microphone picks up again, which helps on calls. Noise suppression lowers steady background noise such as a fan, and can also cut quiet speech. Both start on. Turn them off to see what the raw microphone delivers; changing a switch restarts the test, and the browser may apply it differently or ignore it. After you allow access, the Microphone list shows every input the browser reports, and choosing another one restarts the test with that device. Names appear only once permission is granted, because browsers hide them before. If a device is plugged in later, the list updates.
It measures what the browser hands to a web page, which can differ from what a call app or a recording program receives, because those may use other settings. It does not measure sound quality, hum, room echo or latency. The level is relative to full scale, so it is not a sound pressure level in real-world decibels. The 5-second clip is recorded by your browser in the format it supports, usually WebM or MP4, and the page states which one it used. The clip stays in this tab's memory and is gone when you close or reload the page; nothing is saved, stored or uploaded.
It asks only after you press Start. If you blocked this site before, the browser skips the prompt: click the lock icon in the address bar, set Microphone to Allow, and press Start again. A page opened over plain http:// is not allowed to use the microphone at all.
The page is getting silence. Check that the microphone is not muted (some headsets have a switch on the cable), that the right device is chosen in the Microphone list, and that another app is not holding it. After about half a second of readings the verdict says No signal if the loudest peak stayed below -70 dBFS.
On this meter, speaking at a normal distance should push the level above -45 dBFS while peaks stay under -1 dBFS. Below -45 the verdict says Very quiet; at -1 dBFS or higher in two or more readings it says Clipping. These numbers are this tool's own cut-offs, not an industry standard.
Clipping happens when the sound is louder than the largest value a sample can hold, so the top of each wave is cut off and the voice sounds harsh. The verdict says Clipping when peaks reach -1 dBFS or higher in 2 or more readings. Move back from the microphone or lower the input volume in your system sound settings.
Yes. After you press Start and allow access, the Microphone list shows the inputs your browser reports, such as a built-in microphone, a USB microphone or a headset. Choosing one restarts the test with it.
No. The audio is processed in your browser and not uploaded. The level is calculated on your device, and the 5-second clip stays in this tab's memory until you leave or reload the page. The page counts visits and presses of Start with Google Analytics, never audio.