Online sound level meter

Works right now. No app, no account, no catch.

SoundMeter showing 52.4 dB level, waveform, frequency spectrum and session stats

Actual interface shown after microphone access.

Requires microphone access. Audio stays on your device.

Why we built this

I tried a bunch of online sound meters and noticed something odd - my readings were always lower than what my phone's built-in meter showed. Turns out, browsers apply three layers of audio processing by default: automatic gain control (compresses loud sounds down), noise suppression (removes what it thinks is background noise), and echo cancellation. These are great for voice calls. They're terrible for measurement.

With AGC on, a 75 dB room and a 55 dB room can show nearly identical readings because the browser is normalizing everything toward "comfortable speech level." You're not measuring the room anymore - you're measuring the browser's opinion of what the room should sound like.

This tool explicitly requests that all three processors be disabled via the getUserMedia constraints API. Not every browser fully respects this - Chrome does reliably, Firefox mostly does, Safari is inconsistent. But even partial compliance gives you something closer to the truth than the default.

That's it. That's the whole reason this exists. One problem, one fix.

How the measurement works

Calls getUserMedia() with AGC and noise suppression disabled. Feeds into a 2048-point FFT AnalyserNode. Computes RMS each frame and converts to dB. Some browsers don't fully honor the disable requests.

Limitations and privacy

No calibration - absolute values may drift 3-8 dB. Relative comparisons are reliable. Not for compliance. Audio stays in the browser - static site, no backend. Check the Network tab to verify.

Noise level reference

30 dB is a quiet bedroom - you hear your own breathing. 50 dB is a typical office with a computer running. 60-65 dB is two people talking across a table. 70-75 dB is a busy restaurant where you raise your voice. 85 dB is heavy traffic or a lawn mower - OSHA says limit exposure to 8 hours. 100 dB is a loud club - about 15 minutes before risk. Above 110, you're looking at minutes at most.

Frequently asked

How accurate is this?
It depends on your microphone. We disable AGC and noise suppression to get unprocessed audio, then calculate dB from the RMS amplitude. On a decent laptop mic, readings are consistent enough to compare rooms or track changes over time. But this isn't a calibrated instrument - don't use it for anything where exact numbers matter legally or medically.
Does it upload my audio?
No. The mic stream goes into the browser's Web Audio API AnalyserNode and stays there. No server receives it. You can confirm this yourself - open the Network tab in DevTools while measuring and you'll see zero outbound audio requests. The site is static files on a CDN.
What noise level is safe?
Generally, under 70 dB is fine for extended periods. OSHA sets the workplace limit at 85 dB for 8 hours. Above that, risk goes up fast - at 100 dB you're looking at roughly 15 minutes of safe exposure. These are guidelines, not guarantees. If something sounds uncomfortably loud, it probably is.
Can I use this in a classroom?
That's what the Classroom page (/classroom-noise-meter) is for - an online noise meter for classrooms. Open it on a projector and students see a large number with color feedback. Whether it actually helps depends on your class and how you introduce it. It measures volume only - no speech is recorded or stored.
Is it free?
Yes. Everything works without paying or signing up. There's no premium version. We don't show ads during measurement. The tool is a static website - there's no subscription infrastructure to charge you with even if we wanted to.
What about Zoom/Teams calls?
Run the meter from your normal calling position without talking. If ambient noise is under 35 dB, you're in good shape. Between 35-50 dB, video call software can usually filter it. Above 50 dB, people on the other end will hear it regardless of software noise suppression.