🗣Echo test

✓ Tested & verified Updated: Aug 11, 2026 How the test works

Echo test

Record a few seconds and hear them straight back. The clip is held in this tab only and thrown away the moment it finishes playing.

Loudest peak in the clip

—dBFS

Say a normal sentence, then listen. Hearing yourself back is the only test that catches the faults a meter cannot name: room echo, a fan, a loose cable crackle.

Clip length—
Average level—
Background noise—
Clip statusempty

Record for

Record the raw signal, with the browser's clean-up switched off

Try also

How do I hear what my microphone actually sounds like?

The echo test records five seconds from your microphone and plays it straight back, which is the only way to hear yourself the way other people hear you. A level meter proves the microphone is alive and says nothing about the sound: room echo, keyboard noise, a fan two metres away and the boxy tone of a laptop’s built-in array are all invisible on a meter and obvious in playback.

Listen for three things. Whether your voice is clearly louder than the room; whether the tail of each word rings on after you stop, which is reflection off bare walls; and whether a steady hiss sits under everything, which is gain compensating for a microphone too far away. The recording exists only inside the page and is discarded the moment playback ends, so there is nothing left to delete afterwards.

Playback is a diagnosis, not a verdict. Almost every complaint people get about their audio maps to one of the sounds below, and each has a different fix — which is why hearing the recording is worth more than reading a specification.

What the playback sounds like	What it usually is	What to change
Hollow, with a tail after each word	Reflections off bare walls, glass and a hard desk	Get closer to the mic; add soft surfaces behind you
A steady hiss under the voice	Gain raised to compensate for distance	Halve the distance instead: 40 cm to 20 cm adds about 6 dB of voice and no hiss
Thumps and key clicks	A built-in array picking the chassis up through the case	A headset or external mic; type between sentences, not through them
Your voice quieter than the room	Distance, plus a fan or an open window	Move the microphone first, then the noise source
Clean but boxy and nasal	Beamforming and noise suppression in a laptop array	Nothing on the laptop — this is the array’s tuning, not a fault
A rising squeal	The microphone is hearing the speakers	Headphones, or lower the playback volume

Once the recording sounds right, the number behind it is on the mic test, which shows the peak level in dBFS and the band it should land in. If the playback itself is what sounds wrong rather than the capture, the fault is on the output side — the speaker test and the headphone test separate those.

Frequently asked questions

Why does my voice sound different in the recording?

Because you have never heard it before. While you speak, sound reaches your inner ear twice: through the air, and through the bones of your skull, which carry low frequencies far better than air does. Recording removes the bone path entirely, so playback sounds thinner and higher than the voice in your head. Everyone else has only ever heard the recorded version, which makes it the accurate one.

Why is there hiss or an echo in the playback?

Echo has two sources and they need different fixes. If the microphone can hear your speakers, it re-records the playback and the loop builds — headphones remove that completely. If the ring follows your own voice instead, it is the room: bare walls, a glass desk and an empty corner all reflect. Hiss is different again, and it is almost always input gain raised to compensate for distance.

Why is the recording quiet when the meter looked full?

The meter and the playback measure different things. A meter follows peaks, so a few loud consonants can fill it while the average level of your speech stays low, and average level is what loudness sounds like. Check the output side too: playback runs through the system volume and the current output device, so a recording can be perfectly loud and still play back into a muted or unselected speaker.

Why do Teams and Discord sound worse than this test?

Because a call compresses what this page plays back intact. Meeting software encodes voice to a low bitrate and drops bandwidth when the network tightens, and on a wireless headset the microphone forces Bluetooth into the mono hands-free profile, which is narrow band and audibly worse. This test plays back the captured stream directly, so it is the best case; a call is the same voice through a much smaller pipe.

Is the recording sent anywhere?

No. The audio is held in memory as a blob inside the page, played back from there, and released when playback ends. Nothing is written to disk, nothing is uploaded, and no request leaves the page while the test runs. Stopping the test or closing the tab releases the microphone, and the browser’s recording indicator goes out with it.

My laptop recording sounds thin and boxy. Is the microphone faulty?

Probably not. Built-in laptop arrays are tuned for intelligibility on a call, not for tone: beamforming narrows what they hear, and aggressive noise suppression strips the quiet parts that give a voice body. The array also sits in the chassis, so it picks up the keyboard and the fan through the case. A headset or any external microphone changes this more than any setting will.

How the test works

The page opens the microphone with getUserMedia and records the stream with MediaRecorder into memory — a blob held by the page, never a file on disk. Playback attaches that same blob to an audio element through an object URL, which is revoked when playback ends; the blob is released with it, so the recording is gone the moment you have heard it. Nothing is uploaded and no request leaves the page while it runs. What you hear is the stream after the browser’s own echo cancellation, noise suppression and automatic gain control, because those run on the capture side before any page can reach the samples. That is deliberate rather than a limitation: it is the same processing a call applies, so the playback previews how you will sound in a meeting rather than documenting the room. Stopping the test releases the microphone and the browser’s recording indicator goes out.

Worked example

With loudest sample in the clip 0.5 of full scale, average sample across the clip 0.06, this page works out a peak of −6.0 dBFS. This is the level broadcast desks aim for: peaks between −12 and −6 dBFS are loud without ever touching the ceiling. Leave the gain where it is.

Peak level−6.0 dBFS
Average level−24.4 dBFS
Peak above average18 dB

Sources:

Why playback beats a meter

A meter measures level; ears measure everything else. Reflection, background noise and the colour of a built-in array all sit outside what a single number can show — and all three are what people on the other end actually complain about.

Bone conduction is why the first listen is a shock. Your skull carries the low frequencies of your own voice straight to the inner ear, so the version in your head has a bass register nobody else has ever heard.

Five seconds is enough. Two sentences at your normal speaking distance reproduce every problem worth finding; a longer take mostly reproduces them again.