What "color" means for noise
In audio, "noise" is a random signal, and its "color" describes how that randomness is distributed across frequencies. A noise signal is pure in a statistical sense, not in a pitched sense: there is no musical fundamental, no chord, no melody. What differs between colors is the shape of the spectrum — how much energy sits in low versus high frequencies on average.
The names come from a loose analogy to light. White light is an even mix of visible wavelengths, and white noise is an even mix of audio frequencies. Pink noise has more "warmth" at the bottom of the spectrum, like light skewed toward the red end. The analogy is imperfect — there is no literal wavelength mapping — but the vocabulary stuck.
White noise
White noise has approximately equal energy per hertz across the audible range. On a flat display it looks like a thick horizontal band; on an A-weighted meter it still sounds bright because the ear is more sensitive to mid and high frequencies. Use cases include: speaker and room testing where a broadband reference is needed, masking unwanted sounds when concentration matters, and creating a neutral background in home studios. White noise is the simplest to generate — each sample is an independent random value — and is the common ancestor of all the other colors.
Pink noise
Pink noise has equal energy per octave, which means energy falls by about 3 dB per octave as frequency rises. Because the ear groups frequencies roughly into octaves, pink noise tends to sound balanced and natural — less harsh than white noise but not especially muffled. It is the industry default for room tuning and loudspeaker calibration. When a live-sound engineer plays a reference signal through a PA and walks the room with a measurement microphone, pink noise is typically the signal they use. It is also a popular choice for sleep and focus audio because it is easier on the ears than white noise over long listening sessions.
Brown noise (also called brownian or red noise)
Brown noise falls roughly 6 dB per octave, giving it a deep, rumbling character reminiscent of a distant waterfall or heavy rain behind glass. The name is not a reference to color at all — it comes from Brownian motion, the random walk of particles in a fluid. A running integral of white noise produces a brown-noise-like signal, which is why many implementations describe it as a smoothed or integrated random walk. People who find white or pink noise too bright often prefer brown noise for relaxation, tinnitus relief, or ambient backgrounds.
Blue noise
Blue noise is the mirror image of pink: it rises about 3 dB per octave, emphasizing high frequencies. It sounds hissy and airy rather than deep. Its most widely cited application is in digital signal processing — specifically, dithering. When an audio signal is converted from higher to lower bit depth, adding a small amount of blue noise as dither distributes the quantization error away from the low- and mid-frequency ranges where the ear is most sensitive, so the noise that remains is less perceptible.
Violet (purple) noise
Violet noise rises about 6 dB per octave, making it the high-frequency counterpart to brown noise. Its energy is concentrated at the very top of the spectrum, producing a thin, sharp hiss. Violet noise is used in specialized applications such as certain kinds of tinnitus therapy, as a test signal for high-frequency transducers, and as a building block in more complex audio processes. It is also the derivative of white noise in the discrete-time sense, which is why many software generators implement it by differentiating white samples.
Grey noise
Grey noise is designed to sound uniform across the spectrum to a human listener rather than flat on a measurement meter. It applies an equal-loudness contour so that frequencies the ear is more sensitive to are attenuated, and frequencies it is less sensitive to are boosted. The result is a noise that feels evenly weighted at any point in the audible range. Grey noise is common in psychoacoustic experiments and hearing tests where perceptual uniformity matters more than mathematical flatness.
Picking a noise for a task
- Room and speaker calibration: start with pink noise. It approximates the average long-term spectrum of music and is what most calibration guides assume.
- Focus and concentration: pink or brown noise usually feel less fatiguing than white over long periods. Personal preference varies, so experiment.
- Sleep: brown noise is a common pick for heavier masking and a "warm" feel.
- Infant sleep masking: any steady low-volume noise can help, but keep volume conservative — long exposure at loud levels is not recommended.
- Audio testing: white noise is the simplest reference for verifying a broadband signal path; pink noise is better for frequency-balance work.
- Dithering and DSP experiments: blue or violet noise are the usual high-frequency tools.
- Hearing tests and perceptual experiments: grey noise is the most perceptually uniform baseline.
How volume affects everything
Most "which noise should I use" questions are actually loudness questions in disguise. Any noise becomes harsh at high volume and pleasant at low volume. Calibrate with your ears: start quieter than you think you need, then raise the level until the noise just masks what you want to mask. For long listening sessions, a level you can comfortably hold a conversation over is usually a reasonable ceiling.
Try it
You can generate each of these noises directly in your browser using the Harmonoise Noise Generator, and export them as WAV files at common sample rates and bit depths for use in a DAW, a sleep playlist, or a measurement chain.