A level meter can prove your mic is alive, but it can’t tell you what you sound like. The only check that answers that is a mic test recording: capture a few seconds of your own voice, play it back, and listen the way the person on the other end of the call will.

This site has the whole loop built in — live meter, waveform, then a local-only recorder. Here’s how to run it and what the playback is actually telling you.

Run the test

1. Open the mic tester and press Start microphone test, then allow access.
2. Watch the meter while you say a sentence — healthy speech peaks between −20 and −6 dB. Slamming red means clipping; a flat line means signal isn’t arriving (start with the not-working checklist instead).
3. Press Record a clip, talk for 10–15 seconds, then stop.
4. Play it back right there in the page — or download it if you want a second opinion on other speakers.

The clip is encoded by your browser’s MediaRecorder (usually Opus in WebM) into a local blob. Nothing uploads — there is no server here to receive it — and closing the tab deletes it.

What to say

A good test script deliberately stresses the mic. Twenty seconds is enough:

• Everyday speech at normal volume — “Testing, one two three — can you hear me?” — to set the baseline level.
• Plosives: “Peter Piper picked peppers.” Hard p/b sounds hit the capsule with a puff of air; they expose a mic that’s too close or missing a pop filter.
• Sibilance: “Six sleek swans.” Sustained s sounds reveal harshness and over-bright pickups.
• Quiet then loud: drop to a near-whisper for two seconds, then speak firmly. Quiet passages expose the noise floor; louder ones expose clipping and auto-gain pumping.
• Three seconds of silence — don’t skip this. What you hear in the gaps is the room and the mic’s self-noise, not your voice.

The playback checklist

Listen with headphones if you can — laptop speakers mask exactly the problems you’re hunting.

• Noise floor. In the silent gap, is there hiss (white noise — often preamp gain or a cheap capsule), hum (a 50/60 Hz drone — electrical interference or a ground loop), or computer fan rumble? Light steady hiss on a laptop built-in is normal; a loud hum isn’t.
• Clipping and crackle. If loud syllables crunch or the waveform showed flat, squared-off tops, the input is overloaded. Back off a few inches or drop OS input gain — more mic is not better.
• Plosive thumps. Each ‘p’ shouldn’t arrive like a drum hit. If it does, move the mic slightly off-axis (to the side of your mouth) before spending money on a pop filter.
• Volume. Playback should be comfortably audible without maxing your volume. Speech that sat at −20…−6 dB but plays back whisper-quiet usually means OS-level gain or a headset stuck in a low-quality mode.
• Channel balance. On a stereo input, your voice should sit in the middle, equally in both ears. One-ear playback or a hard left/right image means a mono-on-stereo mismatch or a lopsided headset — the info grid’s channel count settles which.
• Processing artifacts. Warbly, robotic, or “underwater” speech is usually echo cancellation or noise suppression working too hard — the tester reports both flags so you can see what’s active. Bluetooth headsets are the usual suspects: in call mode they switch to a narrow-band profile that no settings will polish.
• Dropouts. Gaps or stutters in the recording point to a busy USB bus, a Bluetooth link fighting interference, or a CPU spike — not the mic element itself.

When the meter and the recording disagree

The meter measures the raw stream the browser hands over; the recording is what survives encoding and playback. If the meter looked healthy but playback is thin or muffled, the suspect list narrows fast: browser-side processing (check the EC/NS/AGC flags on the page), a headset in hands-free mode, or hardware that genuinely sounds like that — USB mic vs headset vs built-in is the honest comparison of what each type delivers.

One more habit worth building: re-record after every change. Lowered the OS gain? Moved the mic? Switched from the Bluetooth headset to the laptop mic? Capture another 10 seconds immediately and compare — recordings are cheap and your memory of “how it sounded before” is not. Two clips side by side settle arguments that an hour of tweaking can’t.

And if the recording reveals a problem you can’t place, two paths cover most cases: the ordered fix checklist for signal-level problems, and sample rate, bit depth & channels for what the numbers on the info grid mean. The permissions model explains why the browser asks before any of this can run at all. Testing ahead of a meeting? Zoom, Teams & Meet each hide their own mic test — and Discord’s quirks get a dedicated check. If what you heard was your own voice arriving late rather than anything wrong with the clip, that’s monitoring latency, not the mic.

Frequently asked questions

How do I test my microphone by recording myself?

Open the tester, allow mic access, then press Record a clip. Say a sentence at normal volume — include a few 'p' and 's' sounds — stop, and play it back in the page. That playback is the closest thing to hearing what a call would transmit, processing included.

What should a good mic test recording sound like?

Your voice clearly louder than the room, no hiss or hum underneath, no crackle on loud syllables, and — on a stereo mic — the same signal in both ears. Speech that peaked between −20 and −6 dB on the meter should play back clean; if it doesn't, the bottleneck is downstream (OS gain, headset mode, or the mic itself).

My recording sounds distorted but the meter never hit red — why?

The meter shows what reaches the browser; playback adds the rest of the chain. Over-aggressive noise suppression can make speech sound watery or robotic, a Bluetooth headset in the wrong mode can drop everything to call-quality mono, and OS-level gain can distort even when our meter looks happy. Compare the waveform (squared-off tops = clipping) with what you hear.

Is the test recording uploaded or stored anywhere?

No. The clip is written to a blob inside this browser tab by MediaRecorder — the site is static files with no endpoint to receive audio. Playback, the download link, everything happens locally; close the tab and the clip is gone.

Why does my recording play in one ear only?

That usually means a mono source feeding one side of a stereo channel, or one side of a headset being tested. Check the channel count in the tester's info grid — '1' means mono is expected, and a single-ear playback of a mono file is normal on some setups. Two channels with sound on only one side is a real imbalance worth investigating.

How long can a test clip be?

The recorder caps at 60 seconds — enough to judge noise, level and tone without producing a file you'd accidentally keep. For a quick check, 10–15 seconds of natural speech is plenty.