Mic Test
This mic test measures your microphone’s input level and background noise in dBFS from the live audio, records five seconds you can play back, and counts how many words of a test sentence your browser’s speech recognition understands. Free, in your browser, no sign-up; nothing leaves your device except the speech-check audio, which your browser sends to its maker’s speech service.
Step 1 · Level and background noise
Step 2 · Record and play back
Step 3 · Does speech recognition understand you?
The quick brown fox jumps over the lazy dog.
The speech check uses your browser’s speech recognition; Chrome, Edge and Safari send that audio to the browser maker’s speech service. The level meter and the 5-second recording stay on your device.
How to use the mic test
- Start the test and allow the microphone when the browser asks. Stay quiet for a second and a half: that window measures your background noise.
- Say a sentence at your normal volume when the prompt changes. The meter shows the live level; the result compares your loudest moment with the background and tells you whether the mic is too quiet, too loud or fine.
- Record five seconds and play them back. This is the only way to hear hum, echo or a muffled headset that the numbers cannot show.
- Run the speech check: say the sentence on screen and see how many of its nine words your browser recognized. A mic can pass the level test and still lose words if it is far away or the room echoes.
How it works
Everything in step 1 comes from the audio samples the browser hands us through the Web Audio AnalyserNode: 2 048 samples per frame, each between −1 and 1. For every frame we compute the RMS (root mean square) and convert it to dBFS, decibels relative to digital full scale: 20 × log10(RMS), clamped at −100 dBFS for digital silence. We use the plain mathematical RMS, under which a full-scale sine wave reads −3.01 dBFS; the AES17 convention would show it 3 dB higher. The meter bar spans −60 to 0 dBFS. “Peak” is the largest single sample seen since you pressed start.
Clipping is our definition, not a standard: a frame clips when at least 1 % of its samples sit at or above 0.99 of full scale. One stray sample is ignored; a run of them means the signal has nowhere left to go and the recording is distorted. Background noise is the 10th percentile of the frame levels measured during the quiet second and a half, so a cough or a chair creak does not count. Your voice is the loudest frame during the four-second speaking window.
The verdict uses two thresholds we chose for speech recognition, not a published standard: your voice should reach −30 dBFS at least once and the background should stay at or below −45 dBFS. Typical speech into a laptop or headset mic at normal gain lands between −25 and −12 dBFS; a quiet room on the same mic sits between −60 and −45. Clipping is reported first, because a clipped signal passes the loudness test while sounding bad. The result also shows the gap between voice and background in dB; when it is 20 dB or more, recognition tends to work even if the background is above our line.
The device lines come from the audio track’s settings (channels, echo cancellation, noise suppression, automatic gain control) and the audio context’s sample rate. “Not reported” means the browser did not expose that value, not that the feature is off. The microphone list comes from enumerateDevices, which only shows names after you have allowed the mic. The five-second clip is made with MediaRecorder in WebM/Opus where the browser supports it and MP4 otherwise, and is kept in the page only. The speech check uses the browser’s Web Speech API, which “involves a server-based recognition engine”, and counts the words of the test sentence that came back in order. Method v1.0, 21 October 2026.
Reference: what the dBFS readings mean
The bands below are our reading of typical laptop, phone and headset microphones at default gain; they are a guide for interpreting the meter, not a specification.
| Frame level (RMS, dBFS) | What it usually is | What to do |
|---|---|---|
| −100 to −70 | Digital silence: muted mic, wrong device selected, or the browser has no signal | Check the system mute switch and the device list; try another microphone |
| −70 to −55 | A quiet room: good background for recording and recognition | Nothing; this is where the background should sit |
| −55 to −45 | A normal room with a fan, traffic or a laptop fan nearby | Still fine for recognition if your voice reaches −30 or louder |
| −45 to −30 | Loud background, or a voice that is too far from the mic | Move closer, use a headset, close the window; raise input gain |
| −30 to −12 | Normal speech at a sensible distance | Nothing; the target zone for calls and voice practice |
| −12 to −3 | Loud, close speech or a high input gain | Fine unless clipping is flagged; lower the gain if it is |
| Clipping flagged | 1 % or more of a frame at full scale: distorted audio | Lower the gain or move away from the mic |
| Verdict | Condition | What it means for voice practice and calls |
|---|---|---|
| Your mic works | Voice peak ≥ −30 dBFS and background ≤ −45 dBFS, no clipping | Speech recognition should hear you; move on to the recording and speech checks |
| Too quiet | Voice peak below −30 dBFS | Recognition will miss words; gain or distance is the fix |
| Noisy background | Background above −45 dBFS | Works when your voice is well above the noise; otherwise fix the room or the mic |
| Too quiet and noisy | Both conditions | The device is usually the problem; try a headset |
| Too loud: the signal clips | Clipping during the speaking window | Loud enough but distorted; lower the gain |
What to do with the result
Your mic works and all nine words came through. The recorder will hear you, so go straight to practice: the English speaking test scores a two-minute answer. In the Fluently app the spoken English assessment is free after installing, and the tutor scores pronunciation, fluency, vocabulary and grammar during a conversation; regular practice is part of a subscription.
Too quiet. Distance is the usual cause: 10–20 cm from a headset boom or a laptop edge is normal, an arm’s length is not. Then raise the input level in your system sound settings (Windows: Sound settings → Input → Volume; macOS: System Settings → Sound → Input). If the list shows a headset and a built-in mic, pick the headset and run again; built-in mics on closed laptops and tablets lying flat often read 10–15 dB lower.
Noisy background. Look at the gap first: a voice 25 dB above the background is understood even in a busy room. If the gap is small, move the fan, close the window, or switch to a headset with the capsule near your mouth. The device lines tell you whether the browser is already applying noise suppression; if it says Off, your system or headset software may have its own switch.
Too loud: clipping. The mic is fine; the gain is not. Lower the input level, turn off any “microphone boost” option, or move the boom to the side of your mouth instead of in front of it. Clipping matters more for recordings and pronunciation scoring than for calls, because a scorer compares the shape of the sound, and a clipped vowel has the wrong shape.
The level passed but the speech check missed words. That is recognition, not hardware: speak at a conversational pace, keep the sentence in one breath, and stay in Chrome, Edge or Safari, which have speech recognition built in. If words still drop, record the five-second clip and listen for echo or a muffled headset. For the sounds themselves, the pronunciation practice guide covers the English sounds that recognizers most often mishear.
Limits and when not to use this test
- dBFS measures how much of the digital range your mic uses, not how loud the room is in decibels SPL. Two microphones in the same room read differently, and the same mic reads differently after a gain change. Do not compare numbers between devices.
- The level you see is after the browser’s processing. Echo cancellation, noise suppression and automatic gain control are applied before we measure, so a mic that reads −20 dBFS here may behave differently in an app that turns those off.
- This is a level, recording and recognition check. It does not measure frequency response, latency, stereo balance or distortion below the clipping threshold.
- The speech check depends on the browser: Chrome, Edge and Safari have speech recognition; Firefox and many in-app browsers do not. A failed speech check in those browsers says nothing about the microphone.
- If the browser has no microphone permission for this site, the test cannot start. Look for the microphone icon in the address bar, or check the site permissions in the browser settings.
- The result is a snapshot of four seconds. A fan that starts later or a voice that trails off at the end of a long answer is not caught; repeat the test if your setup changes.
Frequently asked questions
Why does the mic test show a level but people cannot hear me on calls?
The browser and the call app may be using different devices. The device list here shows which microphone the browser opened; check that the same one is selected in the call app and in the system sound settings. A second cause is a muted track inside the app, which the system meter does not show.
What dBFS level should my microphone be at?
For speech, aim for peaks between −30 and −12 dBFS on this meter with the background at or below −45 dBFS. That leaves headroom before clipping at 0 dBFS and a gap of 15 dB or more above the noise, which is what speech recognition needs. These are our working thresholds, not a standard.
Why is my mic so quiet?
In order of likelihood: the mic is too far away, the input level in the system sound settings is low, the browser picked a built-in mic instead of the headset, or automatic gain control is off on a mic that needs it. Run the test after each change; the voice peak in the result tells you how much each step gained.
Does this mic test record or upload my voice?
The level meter and the five-second clip stay on your device; the clip lives in this tab and disappears when you close it. The speech check is different: it uses the browser’s own speech recognition, and Chrome, Edge and Safari send that audio to the browser maker’s speech service to turn it into text. Nothing is sent to us.
Why does the speech check not work in Firefox?
Firefox has no built-in speech recognition for web pages, so the speech check cannot run there; the same is true of many in-app browsers inside social and messaging apps. The level meter, the recording and the device information work in Firefox. For the speech check, open this page in Chrome, Edge or Safari.
Does the mic test work on a phone?
Yes, in Safari on iOS and Chrome on Android. Allow the microphone when the browser asks, and hold the phone as you would during a call or a voice lesson, because the result depends on the distance. iOS keeps the device list short and may name the mic simply “Default”; the measurements are not affected.
