Audio engineering
How the signal is made
The app synthesizes noise sample by sample on your device. No recording is streamed, repeated, or captured from your microphone.
Open the playerGeneration off the interface thread
An AudioWorklet runs the noise generators on the browser’s real-time audio rendering thread. Interface updates, sheets, and animation do not generate samples, which keeps React work away from the sound path.
Long-period pseudo-random samples become white, pink, or brown noise.
High-pass and shelf filters shape warmth and depth without replacing the source.
Play, pause, volume, timers, and sound changes use smooth ramps.
A large FFT reads the actual outgoing digital signal on a logarithmic frequency axis.
What the analyzer measures
The spectrum is measured in digital full-scale units (dBFS) inside the audio graph. It cannot tell you the physical sound pressure level at your ears. That depends on your device, amplifier, speaker or headphones, distance, and room.
Designed for continuity
The random generator has an extremely long state period and needs no loop buffer. Brown noise uses a leak-controlled integrator and a DC-blocking stage; pink noise uses a stable multi-pole approximation. Output gains are conservative, and every user-facing transition is ramped to avoid clicks.