experiment 29 · audio context · no permissions
your speaker
has a fingerprint.
The Web Audio API can render audio offline — silently, at zero volume, with no hardware output.
Microscopic differences in your audio hardware's signal processing produce a mathematically
unique waveform that identifies your device.
Computed with your volume at zero. No microphone. No sound.
Your digital-to-analog converter hardware produces a mathematically unique output — readable by any website.
Used in production by advertising networks since 2016.
—sample rate
—buffer sum
—hash
OfflineAudioContext is unavailable in this browser.
This API is supported in all modern browsers. If you see this message, try Chrome, Firefox, or Safari.
Audio fingerprinting still works even if your speakers are muted or disconnected.
rendered waveform — computing...
audio fingerprint
rendering...
computed from OfflineAudioContext render — no microphone, no sound output
▶ how does silent audio fingerprinting work?
The OfflineAudioContext renders audio into a memory buffer rather than to speakers. The test signal: a triangle wave oscillator at 10,000 Hz routed through a DynamicsCompressorNode with specific settings (threshold: -50 dBFS, knee: 40 dB, ratio: 12:1, attack: 0 ms, release: 0.25 s). The compressor's nonlinear behavior exposes subtle hardware differences.
The Float32 sample buffer is summed and hashed. Different CPUs, audio subsystems, and audio driver implementations produce slightly different floating-point results due to differences in math co-processor behavior, rounding modes, and DSP instruction sets. The fingerprint remains stable across browser restarts, private windows, and cleared cookies.
This technique was documented by Mowery & Shacham (2012) and later by Englehardt & Narayanan's Princeton Web Census (2016), which found it deployed on over 1,400 of the top 100,000 sites. Brave Browser adds deliberate noise to the audio context to defeat it.