Skip to content
UniKit

White noise generator

Generate white, pink and brown noise in the browser: silent by default and only audible after you press play, with adjustable duration, sample rate and volume plus live mean, variance, RMS and autocorrelation stats.

Runs in your browserEvery computation happens in your browser — your data never leaves this device.

Noise settings

White noise is uniform on [-1,1); pink noise uses the Voss–McCartney multi-octave sum (1/f); brown noise is a leaky integral of white noise (1/f²). Playback happens locally through WebAudio and nothing is uploaded.

Noise colour
50%
MuteSilent by default — pressing Play unmutes
0.5 – 60 s

Signal statistics

Playback:Not playing (muted by default)

Set the parameters and press “Generate & analyse” to inspect this noise signal.

What this tool does

  • Mask office or street noise with white noise — the page stays completely silent until you press Play.
  • For sleeping or focus, pink and brown noise are usually easier on the ears: the energy sits in the low end instead of hissing.
  • Run an audio experiment: set the duration to 2 seconds and compare variance and lag-1 autocorrelation across the three colours to see 1/f versus 1/f².
  • Burn in headphones or test a speaker’s noise floor by generating a fixed-length loop on demand.

Example

Input

Colour “White”, duration 2 seconds, sample rate 44100 Hz, then press “Generate & analyse”

Output

Samples 88200; mean ≈ 0.00, variance ≈ 0.33, RMS ≈ 0.58, peak ≤ 1.00, lag-1 autocorrelation ≈ 0.00

White noise is uniform on [-1,1), so the theoretical variance is 1/3 ≈ 0.333. Pink and brown noise correlate neighbouring samples, so their autocorrelation is clearly positive.

Frequently asked questions

Why is there no sound when the page loads?

The tool starts muted, and browsers only allow audio after a user gesture anyway. Pressing Play unmutes and starts the loop; Stop silences it immediately.

How do pink and brown noise differ?

White noise has equal energy at every frequency (1/f⁰), pink noise falls off as 1/f (more low end), and brown noise falls off as 1/f² (deeper still). The autocorrelation figure makes this concrete: near 0 for white, clearly positive for pink, near 1 for brown.

How is the pink noise computed?

With the Voss–McCartney algorithm: 16 octave rows are kept, and row k is refreshed only every 2^k samples. Summing them gives a spectrum whose power falls off as 1/f, which is what most audio tools do as well.

Does it play forever?

The generated buffer loops. Its length is sample rate × duration, up to 60 seconds; pressing Stop or leaving the page silences it instantly.

Is anything uploaded or stored?

No. Samples are generated in memory and handed straight to WebAudio, the page makes no network requests, and a refresh clears everything.

Keywords:white noisepink noisebrown noisenoise generatorsleep soundvoss mccartney白噪音粉噪音棕噪音助眠噪声生成

Related tools