Audio fingerprinting uses the Web Audio API to process a signal silently and read back the numbers. Nothing plays through your speakers and no microphone is involved.
Where the differences come from
Audio processing is floating point math. Your processor, operating system and browser version can round that math slightly differently, which leaves a small but repeatable pattern in the result.
What else the audio system reveals
- The default sample rate of your audio output
- Output latency and the number of channels
- Which audio features your browser supports
How to check yours
Run the PrintLeaks test and open the Hardware tab to see the audio values your browser reports.