Microphone test: check input levels, record and hear a sample
Admin ·
Start the microphone test, allow browser access, and speak naturally. ToolMellow shows the incoming waveform and digital signal level. To hear the result, record a short sample and use its playback controls.
The live test and sample processing run in your browser. This guide explains device selection, RMS and session peak, negotiated settings, recording limits and cleanup so you can distinguish a useful input check from a physical microphone or call-quality certification.
Start with a live input check
Select Start microphone and respond to the browser permission request. The page requests audio with video disabled through getUserMedia. Nothing is captured automatically when the tool first opens.
Speak a short sentence at your usual distance from the microphone. Watch the waveform and input meter, then make a recording if you also want to hear the captured sound. The live analyser is not connected to speaker output.
A moving meter establishes that the browser is receiving a signal from the selected input. It does not establish which physical capsule is active, whether a meeting app uses the same device, or whether a remote listener can hear you.
W3C — Media Capture and StreamsOnline Mic Test — microphone input check
Resolve permissions and startup errors
Microphone capture requires a supported browser and a secure context such as HTTPS or localhost. Browser permission, operating-system privacy controls and an embedding page’s permissions policy can affect access. Start with the full tool page if an embedded version cannot request permission.
Use the visible error as a starting point, then check the browser’s site permissions, selected input and hardware connection. Error names describe the capture failure reported to this page; they do not by themselves diagnose a broken microphone.
| Condition | What the page reports | Useful next step |
|---|---|---|
| Capture API absent | HTTPS/localhost or browser support is required | Open the secure full tool page in a supported browser |
| Permission denied | Permission was denied | Review site and operating-system microphone permissions |
| No matching device | No matching device was found | Connect an input and check the selected device |
| Device unavailable | The device is busy or unavailable | Close competing capture apps and retry |
| Unsatisfied audio request | Choose System default or another audio input | Change the microphone selection; there is no resolution control here |
| Stream ended | Reconnect the device and start again | Check the connection, then restart capture |
| Other capture error | The device could not start | Recheck the device and browser permissions |
| Recording unavailable | The live test still works | Use the live meter or a browser with recording support |
Choose the input you intend to test
The device selector offers System default and audio inputs the browser exposes. Device names and enumeration can be restricted by permissions or browser policy. A headset microphone, laptop array and USB input can be separate choices.
Changing the selected input during an active test stops this page’s old capture and requests the chosen device. A specific selection uses an exact device identifier, so a disconnected or unavailable input can fail. Try System default or another available audio input; download any sample you need before restarting.
Read the waveform, RMS and session peak
The waveform displays recent analyser samples. RMS describes the sampled signal’s average energy as a digital level; session peak retains the largest absolute sample level observed by this page during the current test. Both are displayed in dBFS, relative to digital full scale.
The analyser uses 2048 samples per reading and the tool applies its own meter scaling. It is not an independent stereo-channel meter, a frequency-spectrum analyser or a calibrated sound-pressure meter. Sample level can reflect browser and device processing as well as the sound reaching the microphone.
| Display or rule | Meaning | Boundary |
|---|---|---|
| Waveform | Recent time-domain analyser samples | Not a frequency spectrum or separate per-channel chart |
| RMS | Average signal energy expressed in dBFS | Not decibels of sound pressure |
| Session peak | Highest sampled absolute peak since this test started | Can remain high after a loud sound has stopped |
| Meter percentage | RMS mapped from -60 to 0 dBFS into 0–100% | Not a microphone quality score |
| Lower display floor | -100 dBFS | Displayed at the floor as ≤ -100 dBFS |
| Near-full-scale warning | A sampled absolute amplitude reached at least 0.999 | A latched warning, not a complete distortion analysis |
Use digital examples without treating them as calibration
For samples x, RMS is sqrt(mean(x²)); the level is 20 × log10(RMS). Peak uses the largest absolute sample before the same logarithmic conversion. The display floors values at -100 dBFS. These formulas explain the readings; they do not convert them to acoustic loudness.
For a constant absolute sample value of 0.25, RMS and peak are approximately -12.0 dBFS. The meter then shows about 80% because it maps a 60 dB interval into a percentage. Real speech varies over time, so this example is a numerical reference rather than a recommended speaking level.
| Numerical sample example | RMS level | Meter implication |
|---|---|---|
| All samples zero | Floor at -100 dBFS | 0% meter; prolonged low signal can show a warning |
Constant absolute amplitude 0.001 | -60.0 dBFS | 0% at the meter’s lower edge |
Constant absolute amplitude 0.125 | Approximately -18.1 dBFS | Approximately 70% |
Constant absolute amplitude 0.25 | Approximately -12.0 dBFS | Approximately 80% |
Constant absolute amplitude 0.5 | Approximately -6.0 dBFS | Approximately 90% |
A sample reaches 0.999 | RMS still depends on all samples | Near-full-scale warning latches for the session |
Interpret the settings the browser reports
After capture starts, the summary shows sample rate in Hz and channel count when reported. The settings table lists echo cancellation, noise suppression, automatic gain control and target latency. getSettings describes the current track settings, rather than an exhaustive list of the hardware’s capabilities.
On and Off reflect reported boolean settings. Not reported means this track did not provide that value; it does not mean zero or a failed microphone. Target latency is a browser setting in seconds, not a measured round-trip delay, network delay or promise about a meeting app.
Account for gain and browser processing
Echo cancellation, noise suppression and automatic gain control may change the signal reaching the analyser or recording. The tool reports available settings but does not offer switches to force those processing modes or a slider to change the operating-system microphone gain.
Compare devices with a similar speaking distance and phrase, and change one external gain or input setting at a time. Avoid treating the loudest meter as the best microphone. A setting shown by this browser does not establish what another app requests or how its audio processing behaves.
Record a sample to hear yourself
While the test is active, select Record up to 10 seconds. Speak, then select Stop recording or let the page’s stop timer run. Once the browser finishes encoding, the local sample appears with playback controls, Download recording and Delete sample.
Use headphones during playback, especially if live capture is still active. The live waveform has no speaker monitoring path, so you hear the recorded sample through playback rather than a continuous live return. Playback also depends on the browser, output device and volume.
Understand the recording timer and size guard
The recording workflow uses MediaRecorder, requests data chunks every 250 ms and schedules a stop after 10000 ms. It rejects a sample if cumulative received encoded data exceeds 10000000 bytes, shown as the 10 MB limit. These are application guards rather than a guarantee of exact duration or total memory use.
Browser scheduling can delay chunks and timer handling. A format reported as supported can still fail during recording. Empty data, an encoding error, an oversized sample or a constructor failure produces an error instead of a playable sample; the live input test can remain available. Retry with a shorter sample or another supported browser.
Download the sample in the browser’s available format
Download recording saves the completed local Blob. The filename begins toolmellow-microphone and its extension follows the recording MIME type: WebM uses .webm, Ogg uses .ogg and MP4 audio uses .m4a. An unrecognized type uses .bin. For example, a WebM sample is toolmellow-microphone.webm.
The page tries supported audio/webm with Opus, audio/ogg with Opus and audio/mp4 before using the recorder’s default when necessary. It does not provide an MP3 export selector or a separate audio transcoder. Check the downloaded file in the app where you intend to use it.
Separate stopping capture from deleting a sample
Stop test releases the media tracks owned by this page and closes its audio-analysis context. If recording is in progress, stopping capture also ends that recording so the browser can finish the sample. A completed sample can remain available for playback, download or deletion after capture stops.
Delete sample clears the page’s sample and revokes its object URL; it does not stop an otherwise active live test. Starting a new test discards the previous sample and resets session measurements. Download an important sample before restarting or leaving the tool. Stopping this page does not stop other apps that independently use the microphone.
Troubleshoot silence and near-full-scale signals
The page shows a low-signal warning after more than 5000 ms without a reading above -70 dBFS. Check the selected input, hardware mute, physical connection and operating-system input gain. This heuristic does not prove absolute acoustic silence or identify a defective microphone.
A sampled peak at or above 0.999 latches the near-full-scale warning until a new test starts. Lower external input gain if playback sounds distorted, then compare another sample. The warning can remain after the current signal falls; neither its presence nor its absence is a comprehensive distortion test.
Know which actions keep or clear the sample
Live capture, displayed settings and a completed recording have different lifetimes. When capture stops, the waveform is cleared and the live level returns to its floor, while the last settings and session peak can remain visible. An idle display is not evidence that capture is still running.
Leaving the tool cleans up its owned capture and sample URLs. A browser reload starts a fresh workspace. A recording that you intentionally downloaded is a separate file and remains under your control outside this page.
| Action | Capture and measurement state | Completed sample |
|---|---|---|
| Open the tool | Idle; no automatic microphone request | None |
| Start a test | Requests the chosen input; resets session values | Previous sample discarded |
| Stop recording | Live capture can continue | Finishes encoding for playback or download |
| Stop test | Owned tracks stop; analysis closes; settings/peak can remain | A completed or finishing sample can remain |
| Delete sample | An active live test continues | Page sample and its object URL cleared |
| Switch an active input | Old tracks stop; chosen input is requested | Previous sample discarded |
| Leave or reload | Page-owned capture is cleaned up | Page sample is not retained; downloaded files are separate |
Understand what stays in the page
The microphone stream is analysed and a sample is encoded locally in the browser; the tool does not upload that audio for analysis. Loading the website and its assets can still use the network. Local audio processing does not guarantee that a first visit works offline.
This tool does not create an account recording library or preserve the page sample across reloads. Download is an intentional save outside page memory; deleting the page sample does not delete a file you already saved. Browser or operating-system permission controls remain available when you want to revoke future access.
Make a useful check before your next call
Begin with the input you plan to use, speak at a normal distance and listen to a short recording. A working local sample helps separate browser input problems from issues in another app, but it does not test a remote participant, call service, network path or speaker quality.
Other microphone-test sites may show additional ratings or metadata. Their labels are not a scoring system implemented here. Use this tool’s actual input, recording and cleanup controls for the local check, then confirm the input and output settings in the application you intend to use.
- Open the secure tool page and choose the intended audio input.
- Start capture and allow the relevant browser permission.
- Speak normally and inspect the waveform, RMS and session peak.
- Record a short sample and play it back through headphones.
- Download a sample you need to keep, then stop the live test.
- Confirm the input independently in your meeting or recording app.
Frequently asked questions
Why is my microphone not detected?
Check the connection, hardware mute, operating-system privacy controls and browser permissions. Try System default or another available audio input. The selected device may have disconnected, and a capture error alone does not establish a hardware fault.
How do I allow microphone access in my browser?
Select Start microphone and respond to the browser prompt. If access was blocked earlier, review the site’s microphone permission and your operating-system microphone settings. Use HTTPS or localhost and a browser that supports capture.
Can I hear myself while testing the microphone?
Use Record up to 10 seconds, then play the completed sample. The live waveform is not routed to speakers, so it does not provide continuous live monitoring. Headphones help prevent playback feedback while capture remains active.
What do RMS and session peak mean?
RMS represents the current sampled signal’s average energy in dBFS. Session peak retains the highest absolute sampled peak during this test. A past loud signal can keep session peak high after the live signal becomes quieter.
Does this microphone test measure sound pressure?
No. dBFS is a digital sample-level reference, not calibrated sound pressure in dB SPL. Gain and processing can affect the readings, and the percentage meter is not a microphone quality score.
How long can I record a microphone sample?
The page schedules a stop after ten seconds and rejects cumulative received encoded data above 10000000 bytes. Browser scheduling can delay the stop or data delivery, so the timer is not an exact-duration or total-memory guarantee.
Why is microphone latency marked Not reported?
Not reported means the browser did not expose that track setting. It does not mean zero latency. A reported target latency is a browser setting in seconds, not a measured round-trip or call delay.
What happens when I stop the microphone test?
The page stops its owned tracks and closes audio analysis. An in-progress recording is ended for finalization, and a completed sample can remain for playback or download. Last settings and session peak can remain visible while capture is idle.
Why does a recorded sample sound different from the live meter?
The meter summarizes digital samples, while playback includes recording encoding and your output device, volume and listening environment. Browser processing can also affect capture. This page does not measure end-to-end acoustic or call performance.
Does Delete sample stop the microphone?
No. It removes the page’s recording and revokes the sample URL. Use Stop test to release active capture. A file you already downloaded is separate and is not deleted by this button.
Will changing microphones keep my previous recording?
Changing an active input restarts capture and discards the previous sample. Download it first if you need to keep it. If the chosen device cannot satisfy the request, choose System default or another audio input.
Can I download an MP3 or choose an audio codec?
There is no MP3 export or manual codec selector. The browser recorder determines an available supported format, and the download extension follows its MIME type. A browser that supports capture may still lack working recording support.
Why does the high-level warning stay after I speak quietly?
The near-full-scale warning latches after a sampled absolute peak reaches 0.999. It resets when a new test starts. Session peak also retains the highest observed peak; neither represents only the current instant.
Does a passing local test prove my next call will work?
No. It checks this browser’s selected input and local sample workflow. Another app can choose a different device or processing setup, and remote audio depends on additional output, network and service conditions.