RunTheTests
/

Audio Fingerprint

See the audio-stack fingerprint your browser produces — computed silently, never played.

How this test works

An oscillator is passed through a dynamics compressor and rendered in an offline audio context — one that computes samples as fast as it can and never sends them to an output device, so nothing is audible at any stage. The resulting floating-point samples differ very slightly between platforms and browser builds, because the signal processing is implemented differently and rounding accumulates through the graph. Summing a slice of them produces a value that is stable on one machine and varies across others. This page computes it, shows it, and discards it.

What the results mean

Audio hash
A digest of the rendered samples. Stable for your platform and browser build.
Samples analysed
How much of the rendered buffer contributed to the value.
Nothing played
The context is offline. No sound reaches your speakers at any point.
Unavailable
The offline audio context is not available, which closes this surface.

Common problems and fixes

Will this make a sound?
No. The audio is rendered offline and never connected to an output device. The context exists only to compute samples.
My value matches my colleague's
Likely, if you have the same operating system, browser and version. Audio has less discriminating power than WebGL or fonts, which is why it is a supporting signal.
Can I block it?
Not easily without breaking real audio, which is what makes it awkward to defend against. Browsers with strong anti-fingerprinting modes normalise it.

Frequently asked questions

Does audio fingerprinting listen to my microphone?

No, and it never has. It generates a signal internally and measures how the browser's own processing renders it. No microphone is involved and no permission is requested.

Why does audio processing differ between machines?

Because the digital signal processing is implemented in the browser and platform code, and floating-point rounding differences accumulate through the graph. Identical instructions produce very slightly different numbers.

Is it a strong identifier?

Weaker than WebGL or fonts on its own — many people share a platform and browser version. It contributes as one signal among many.

More in Privacy