Sleep Sounds — White, Pink, Brown

Sleep sounds play continuous white, pink or brown noise synthesized live in your browser with the Web Audio API, with 15, 30 and 60-minute timers and an automatic 30-second fade-out. Nothing is streamed, nothing is stored, and playback continues if the connection drops.

How to use it

1. Choose a noise colour. Three cards select White Noise (0 dB per octave), Pink Noise (−3 dB per octave) or Brown Noise (−6 dB per octave, the default).
2. Set a timer. The pills select 15, 30 or 60 minutes, or Continuous (Off) for playback that runs until you stop it. The default is 30 minutes.
3. Press Play sleep sounds. The status badge changes to Active sleep timer and the countdown ring shows the time left. A 1.5-second fade-in keeps the start from being abrupt.
4. Trim the level. The volume slider runs from 0.01 to 0.8 linear gain with a dBFS readout, and a warning appears above the 0.45 safety limit (about −6.9 dBFS). Start lower than you think you need.
5. Warm the tone if white noise hisses. The Bedtime Warmth Filter applies a low-pass at 3.5 kHz, rolling off the top end for sensitive ears.
6. Leave it running. Bedside Night View dims the screen to a large bedside clock; tap the overlay to return. Download WAV Loop renders a 10-second stereo file for offline players and flights.

Masking: why a steady noise floor helps

Sleep noise works by masking rather than soothing: a steady broadband floor raises the ambient level in the room, so sudden sounds arrive with far less contrast against the background. A car door or a barking dog wakes people mainly through its suddenness, the step change from quiet to loud, not through absolute level. A smooth, continuous spectrum reduces that step change, which is why the noise can sit at a low, unobtrusive level while still doing the work.

The three colours are one noise source shaped by different filters, not three different recordings. That matters when you pick one: the spectrum, not the mood label, decides which sounds get masked. Equal-loudness behaviour (ISO 226) also means low-frequency energy must be stronger before it sounds as loud as midrange energy, so brown noise at the same level sounds softer and deeper than white.

Noise colour	Spectral slope	Character	Masks best
White	0 dB/octave, equal energy per hertz	Bright hiss, like TV static	Sharp high-frequency transients: horns, clatter, cutlery
Pink	−3 dB/octave, equal energy per octave	Steady rain or wind	Speech and broadband room noise; a middle setting for most rooms
Brown	−6 dB/octave, energy concentrated low	Deep rumble, waterfall, jet cabin	Low-frequency traffic rumble, footsteps, voices through walls

The timer and the fade-out

The fade-out is the point of the timer: a hard cut to silence can wake you, because the auditory system reacts to the change even while you sleep. The engine schedules that fade on the audio clock through AudioParam automation rather than a JavaScript timer, so it stays smooth when the tab is busy or the screen is dark. Every preset ends with a 30-second glide to zero; stopping manually releases the sound in 0.4 seconds instead.

• 15 minutes (900 s): a nap or a short settling period.
• 30 minutes (1800 s): the default; enough time to fall asleep inside the fade.
• 60 minutes (3600 s): slow sleep onset or noisy neighbours.
• Continuous (Off): no fade at all; the badge reads “Playing continuous”.

Keeping the level sane

A practical calibration beats a number on a slider: set the volume so a normal conversation at arm’s length would still be audible over the noise, and keep phone speakers at least a metre from the pillow. If you want the number, run the Decibel Meter at pillow height and aim for the low 50s to about 60 dBA. All-night sessions are the case to be careful with, because a sealed earbud can deliver the same level for eight hours; speakers are the safer choice.

Accuracy, limits and troubleshooting

• The tool synthesizes noise; it does not stream rain or ocean recordings. A “rain” character comes from the pink spectrum, not from a field recording, so do not expect individual droplets.
• dBFS is digital full scale, not dB SPL. The readout describes the signal the page generates; the sound pressure at your ear depends on the device, its volume setting and the distance to your head.
• The volume warning above 0.45 gain is a fixed gain limit, not a calibrated loudness measurement. Treat it as a nudge to check the level, not as a certified safety threshold.
• Offline playback works because the AudioWorklet runs locally. The Download WAV Loop button renders a 10-second stereo loop at your chosen colour and level for players that cannot run a browser tab.
• Background behaviour differs by platform. Desktop browsers keep the audio running in a background tab; on phones, locking the screen can suspend browser audio, so test your device before relying on it all night.
• The timer options are fixed at 15, 30 and 60 minutes, or continuous. There is no custom countdown field; the four pills are the whole set.

Related tools

• Noise Generator — the full noise lab with more colours, filters and level control.
• Brown Note — a dedicated deep brown-noise generator for low-frequency masking.
• Binaural Beats — slow-beat tones for relaxation experiments, which need headphones to work.
• Decibel Meter: measure the level at your pillow so the 60 dBA guideline becomes a real number.