Mic Loopback Test — Hear Yourself

A mic loopback test records a short sample of your voice and plays it straight back, so you hear the signal other people will hear instead of the version filtered by your skull. This page captures up to 10 seconds into browser memory with MediaRecorder and loops it locally. No audio leaves your device.

How to use it

The controls run top to bottom: device and capture mode, record and monitor, then playback, measurement and export.

1. Pick the microphone. Grant permission, then choose your input in the device selector; labels appear once the browser has access. Use the refresh button if you plug in a device after the page loads.
2. Choose a capture mode. Studio Raw disables echoCancellation, noiseSuppression and autoGainControl and captures mono, the honest view of capsule and room. Enhanced DSP enables all three so you can hear how Zoom, Meet or Discord will process your voice. Switching modes closes the open stream, so record again.
3. Press Record 10s Loopback. Speak at your normal working distance. The progress bar counts 0.0 s to 10.0 s and the recorder stops itself at the limit, or you can stop early with the same button.
4. Watch the live meter. The bar is RMS with 125 ms fast ballistics, the marker is peak hold decaying at 12 dB/s, and the scale runs -60 dBFS to 0 dBFS with a -18 dBFS target mark. The clip badge lights at 0.985, about -0.13 dBFS.
5. Listen to the automatic loopback. Playback starts once the capture is decoded, with Loop ON; use Play/Pause, Restart or the loop toggle for a single pass.
6. Scrub the forensic waveform. Click or drag the min/max envelope to move the playhead; the arrow keys step 0.25 s, Shift plus an arrow steps 1.0 s, and Space or Enter toggles playback.
7. Export or re-record. Export WAV saves mic-loopback-recording.wav; Re-record replaces it with a fresh 10-second take.

Direct monitoring and acoustic feedback

The Direct Monitor checkbox routes the microphone through a gain stage straight to the output, which is the fastest way to check that a headset boom is live. It is also the only mode where microphone and speakers are connected at the same time. If the output is a laptop speaker rather than headphones, the microphone hears the speaker, the speaker amplifies the microphone, and once the loop gain passes unity the system howls at whichever frequency first exceeds it, usually in a narrow band around 1 to 4 kHz. That is acoustic feedback, not software echo. The tool gates live monitoring behind the checkbox, labels the control “headphones required”, defaults monitoring off, and releases the capture track as soon as recording stops and monitoring is switched off. The 10-second loopback path never connects the microphone to the output live, so it cannot howl.

Safety: set the volume low before enabling Direct Monitor, wear the headphones rather than resting them around your neck, and never turn the volume up to chase a howl. Feedback builds in milliseconds and can startle you or stress small drivers. Mute the microphone in the conferencing app before testing an open-speaker setup.

Loopback latency arithmetic

Digital delay is buffer size divided by sample rate, and each direction of the trip counts. The table gives one buffer’s worth of delay at common rates.

Buffer (samples)	44.1 kHz	48 kHz	96 kHz
128	2.90 ms	2.67 ms	1.33 ms
256	5.81 ms	5.33 ms	2.67 ms
512	11.61 ms	10.67 ms	5.33 ms
1024	23.22 ms	21.33 ms	10.67 ms

Web Audio renders in 128-sample quanta, so a monitored signal crosses at least an input buffer, one render quantum and an output buffer. With 128-sample buffers at 48 kHz that is about 8 ms before converter overhead; with 512-sample buffers it exceeds 30 ms. Delayed auditory feedback in the 15-40 ms region is where speakers begin to stumble over their own words, which is why recording first and listening afterwards is the better way to judge tone. The “0 ms (Direct)” readout is a mode label, not a measured round trip.

Reading the diagnosis card

After every capture the card reports Peak Level, Average RMS, Crest Factor, Duration and Sample Rate, with an assessment badge. Peak and RMS are in dBFS, and crest factor is their difference on the same buffer.

Badge	Condition evaluated by the code	What to do
SILENT	Peak at or below -45 dBFS	Check the hardware mute, the input level, and that the right device is selected
QUIET	Peak between -45 and -24 dBFS	Raise the microphone input gain in the OS or on the interface
CLIPPED	Any sample at or above 0.995, or a peak at or above -0.1 dBFS	Lower input gain and re-record; this take is digitally distorted
LOUD	Peak above -3 dBFS with no clipped samples	Move back from the capsule or reduce gain
OPTIMAL	Fallback when the peak sits between -24 and -3 dBFS with no clipping	The printed target corridor is -18 to -6 dBFS, marked at -18 dBFS on the meter

Crest factor shrinks as compression, limiting or AGC reduce the distance between the loudest syllable and the average level, so a raw take with plosives shows a wider crest factor than a conferencing stream. The checks run in the order listed, so a very quiet take that also clips reports as quiet; raise the level and record again.

Accuracy, limits and troubleshooting

• Capture is mono at channelCount 1, so a stereo array is summed before this page sees it.
• MediaRecorder compresses the take (Opus in WebM, AAC in MP4); the decoded buffer is close but not bit-exact.
• Peak and RMS are sample-based, so true-peak inter-sample overs are missed; a converter can clip while the sample peak reads -0.1 dBFS.
• Levels are dBFS only; this page has no calibration offset and cannot report dB SPL.
• System processing can override Studio Raw: Windows enhancements, macOS voice isolation and vendor utilities alter the stream before the browser receives it.
• The loopback exposes HVAC hum, keyboard clicks, fan noise and flutter echo that your brain normally filters out, so judge microphone, gain and room together.
• If another application holds the microphone, permission can fail or the stream can arrive silent; close it and refresh.

Related tools

• Microphone Test — Confirm the browser receives any signal at all before you judge tone.
• Mic Quality Test — Compare frequency response and noise across different microphones.
• Audio Latency Test — Measure the round-trip delay of your interface instead of relying on buffer arithmetic.