White, pink, brown and violet noise

The colors are not a marketing vocabulary. Each names a specific spectral slope — how energy is distributed across frequency — and the slope is what you hear as character.

The four

White — flat, 0 dB per octave. Equal energy at every frequency. Because human hearing has far more resolution in the upper range, flat energy sounds bright, even hissy. Excellent at masking, tiring over hours.

Pink — −3 dB per octave. Equal energy per octave rather than per hertz, which is much closer to how hearing is organized. This is the one most people can tolerate longest, and the one that appears most often in the sleep literature. It is the default in this catalog: 522 of 777 sessions.

Brown — −6 dB per octave. Steeply weighted to the low end. Reads as a rumble, distant surf, a plane cabin. Good for sleep and for masking low-frequency intrusion; poor at masking speech, which lives higher up.

Violet — +6 dB per octave. The inverse of brown, weighted to the top. Rarely pleasant on its own, but genuinely useful at low level for adding air, and it is the color most used in tinnitus-related sound work because it can be shaped around a specific region.

Which to choose

• Masking speech — pink or white. Speech energy sits roughly 300 Hz to 3 kHz; brown does not reach it.
• Sleep — brown or a darkened pink, at low level.
• Long focus blocks — pink. It is the compromise that survives hours.
• Bright, thin playback (laptop speakers) — brown or dark pink, to compensate for the low end the speaker cannot produce.

The part that matters more than the color

The catalog names 17 distinct beds — "Brownian", "Deep Brownian", "Sub Pink", "Dark Violet" — from only four generators. The difference is a spectral tilt applied on top: a low-pass filter that darkens the bed by a specified amount.

That is where most of the real character lives. "Pink" and "Dark Pink" are the same generator with very different results.

Those names used to be written by hand, and they drifted. Twenty-one label strings resolved to twenty-six audible beds, and 259 sessions advertised a qualifier — "Soft Pink", "Light Violet" — while specifying a tilt of zero, which is bit-for-bit the plain color. The name is now computed from the generator and the tilt, so a bed cannot be called something it is not.

There is an engineering detail here that is more interesting than it sounds. Filtering out high frequencies also removes loudness, so you have to put the level back — and how much you need depends on the noise's own slope. A correction that is exactly right for white noise overshoots brown by roughly eight times and clips.

So the tilt measures the loss directly rather than assuming a spectrum, tracking input and output level continuously and applying only the gain actually required. The result is that changing a bed's color does not also change how loud it is. Level and color stay independent controls, which is what lets a session specify both precisely.

Level

Every session specifies its own bed level, between 0.02 and 0.2 in the engine's scale. That is a deliberately wide range: some sessions are almost bare tone, others are mostly bed with a tone inside it.

If a session feels too exposed or too buried, that is a property of the session rather than something to fix with your system volume — try a different one rather than fighting it.