White Noise Generator
Most noise pages are players for a recorded loop: a file to fetch first, and a few minutes of audio that comes round again for as long as you leave it on. This one has no recording in it. The browser's audio engine generates fresh samples and schedules them end to end for as long as you leave it running, so nothing repeats and nothing is streamed. Three colours are offered, white, pink and brown, and each is a different slope of the same signal rather than a different file. It opens at 18 per cent, fades in over the time you set, and stops itself after 25 minutes, an hour, 8 hours or a length you type.
of 25 m
Silent: nothing is playing and no audio node is open. Press play to start white noise at 18 per cent.
- Equal energy in every hertz, so the spectrum is flat: the brightest of the three, a hiss.
- Silent. No audio node exists until you press play.
- 1.5 second buffers are generated and scheduled end to end, so the sound never comes round again.
- Silent, and the tab stops using the audio device.
It starts quiet, and the level is yours. The tool opens at 18 per cent, and the slider follows a perceptual curve, so half way along is a quarter of full output. On headphones, which sit against the ear, start lower than you think you need and come up. Every start is faded in over the time you set and every stop is ramped down, so nothing here begins or ends with a click.
Accuracy. The noise is generated in the browser from a random source, so it never loops and nothing is downloaded. White is equal energy per hertz, pink falls about 3 dB per octave and brown about 6, which is why brown sounds deeper. This is a sound generator: it makes no claim about sleep, concentration, tinnitus or any health effect.
Common questions
- What is the difference between white, pink and brown noise?
- It is the slope of the spectrum, and nothing else. White noise has equal energy in every hertz, so it is flat and sounds bright, close to a hiss or an untuned radio. Pink noise falls about 3 dB for every octave you go up, which works out as equal energy per octave, and reads as softer and more like steady rain. Brown noise falls about twice as fast, about 6 dB an octave, which takes most of the top end out and leaves something closer to a waterfall or a distant engine. Here white is drawn one independent sample at a time, pink is built by the Voss-McCartney method of summing rows redrawn at halving rates, and brown is white noise run through a leaky integrator.
- Does the noise loop, and is anything downloaded?
- No, on both counts. There is no audio file on this page and nothing is fetched while it runs. A scheduler fills short buffers with freshly generated samples and books them on the audio clock one after another, with the filter state carried across each join, so what you hear in the tenth minute was generated in the tenth minute. That is the practical difference from a looped recording: there is no point at which the same few minutes start again, and no pattern to notice once you have heard it a few times.
- Does this page make any sleep, medical or tinnitus claim?
- No. This is a sound generator: it produces a signal and describes it by its slope, and that is the whole of what it does. Anything to do with tinnitus, hearing or a medical condition belongs with a clinician rather than with a web page, and a question about a baby's hearing belongs with your paediatrician.
- How loud is it, and why does it open at 18 per cent?
- The level slider follows a perceptual curve: the number is squared to get the gain, so half way along the slider is a quarter of full output and the quiet end of the range is usable instead of being crammed into the first centimetre. It opens at 18 per cent because a page that started at full output would be unpleasant on speakers and worse on headphones, which sit against the ear. Start lower than you think you need and come up. Every start is faded in over the time you set and every stop is ramped down, so nothing begins or ends with a click.
- Can I set it to stop on its own?
- Yes. There are buttons for 25 minutes, 1 hour and 8 hours, an Off setting that runs until you press stop, and a field for a length of your own. Type minutes such as 45, a length such as 1h30m, or a clock time such as 1:30:00, up to 12 hours. The countdown is on the stage while it runs, and when it reaches zero the sound is faded out and the audio nodes are released, the same way the Stop button does it.
- Will it keep going if I switch tabs or lock my phone?
- On a phone, usually not: the browser suspends audio when the screen locks or the tab goes to the background, and the noise stops. On a desktop browser a tab that is making sound is normally left running in the background, but that is the browser's decision and not something a web page can hold it to. Leaving this tab open and visible is the reliable way to run it. Nothing is stored beyond your own settings, in this browser, so the tool opens where you left it.
- Can I save the noise as a WAV or MP3 file?
- Not from this page. The samples exist only while they are playing and are handed straight to the audio engine, so there is no file here to save and no export button. If you need a fixed piece of audio to drop into a video or a sampler, record the output with an audio editor, or generate it there. What this page is good at is running for hours without repeating, which a file cannot do.
The noise is generated in the browser from a random source, so it never loops and nothing is downloaded. White is equal energy per hertz, pink falls about 3 dB per octave and brown about 6, which is why brown sounds deeper. This is a sound generator: it makes no claim about sleep, concentration, tinnitus or any health effect.