A practical white noise generator produces a bounded random signal with a flat expected power spectral density across the audible range, defaults to a 15% tool gain, and stops instantly on demand. White noise is useful for masking irregular environmental sounds, providing a steady reference for audio testing, and giving some listeners a constant backdrop, but only when playback is set up correctly. The Lizely White Noise Generator runs entirely in the current browser tab: it fills a two-second mono Float32 buffer with independent-looking samples uniformly distributed from -1 inclusive to 1 exclusive, then loops that buffer through a single GainNode whose level the on-screen slider controls. Because the slider only governs this page's gain, your operating system mixer, speakers, headphones, and listening room all influence the actual sound pressure you hear. The tips below focus on getting clean, predictable playback from the generator, and the mistakes section calls out habits that turn a useful tool into an uncomfortable or wasteful one.
What good white noise playback actually requires
White noise is defined by an expected power spectral density that is flat across the frequencies under consideration; NI separately describes wide-band noise as flat in the frequency domain, and the W3C Web Audio API specification governs how a generated signal reaches your speakers in a browser tab. A browser generator cannot produce mathematically infinite-bandwidth noise, so the Lizely tool is honest about being a practical, band-limited approximation. Each new playback session requests a fresh 32-bit seed and runs the three-shift xorshift32 recurrence documented in the Journal of Statistical Software, mapping every unsigned 32-bit output linearly into [-1, 1). The samples are not correlated, the bounds hold by construction, and the fixed two-second buffer caps memory use at 384,000 Float32 values — about 1.5 MiB at the maximum accepted sample rate of 192 kHz, or roughly 352 KiB at the more common 44.1 kHz. Playback honors the current browser's native sample rate, accepted only between 8 kHz and 192 kHz.
Because the slider controls a single linear GainNode and nothing else, the tool's output only matches your expectations when you account for everything downstream of the page: the operating system volume, the browser tab volume where applicable, the headphone amp, the headphone fit, and the room itself. A flat, predictable signal in the browser can still be uncomfortably loud through efficient headphones, and a quiet one can disappear inside a noisy room. Good playback starts with that gap in mind, not after the fact.
Use the White Noise Generator in three deliberate steps
The Lizely White Noise Generator exposes only two buttons and one slider, and the operating steps match the controls exactly. Follow the order below the first few times you use it; the rhythm becomes muscle memory after that.
- Set your device output low, then leave the tool at its conservative 15% starting volume or reduce it further. This is the single most important habit for safe, repeatable playback.
- Select Start white noise and adjust the 0–100% slider only as needed. The signal begins the moment the AudioContext resumes, and the buffer starts looping through the GainNode.
- Select Stop before leaving the page or whenever the sound feels uncomfortable. Stop tears down the active audio graph so you never leave a live signal running in a background tab.
If you want a deeper walkthrough of the same workflow, the browser-based white noise guide covers the digital side with extra context.
Common mistakes that hurt your ears or your workflow
Most white noise headaches come from a small set of repeated habits. None of them are subtle once you name them.
- Trusting the percentage as a safety number. The 15% starting gain is a conservative tool default, not a dB measurement. Your device, headphones, and room set the actual loudness. Treat any percentage as a starting point and lower it whenever the sound feels sharp.
- Maxing the slider to "make it work." 100% on the page is not endorsed as safe. If you cannot hear the noise at a comfortable percentage, the answer is almost always to fix your environment — better seal on headphones, different room — rather than to push the gain.
- Clicking Start repeatedly to "boost" the signal. The Start control is guarded while playback is starting or active, and repeated clicks cannot create overlapping players. Pressing Start over and over just queues requests that the same cleanup path will tear down.
- Leaving the page open and walking away. The signal loops indefinitely until you click Stop or navigate away. The same teardown that runs on Stop also runs on component unmount and on startup failure, but you should not rely on background tabs to manage that for you.
- Expecting medical or acoustic guarantees. The tool is a simple controllable signal source. It does not calibrate sound pressure, does not measure hearing, and does not claim to diagnose, prevent, or treat any condition. Use it as a reference signal and a backdrop, not as hearing protection.
- Assuming the noise is Gaussian. The samples are bounded uniform, not Gaussian. Both are "white" when uncorrelated with a flat expected power spectral density, but their amplitude distributions differ. If your downstream test depends on a specific amplitude distribution, account for that.
How the 0–100% slider actually behaves
The slider's job is narrow on purpose. It writes a single linear gain value between 0.00 and 1.00, derived from the displayed percentage divided by 100, into a GainNode that lives on this page. It does not touch the operating system mixer, the browser master volume, your headphone amplifier, or the analog path after the digital-to-analog converter. The table below shows what the slider does and what stays outside its reach.
| Inside the slider's reach | Outside the slider's reach |
|---|---|
| This page's GainNode value (0.00–1.00) | Operating system output volume |
| The audible loudness ratio while the tab is foregrounded | Headphone or speaker sensitivity |
| Whether this page's buffer plays at all (0% mutes this page only) | Background tab playback policy in your browser |
| The default starting gain of 15% on each new session | Any measurement of sound pressure level |
Two practical corollaries follow. First, setting the slider to 0% silences this page's signal but does not unmute anything else, which is useful as a quick mute without changing the rest of your setup. Second, the only honest way to describe the slider is "tool gain, linear, this page only" — a phrasing worth repeating to anyone who asks why 15% sounds different on their phone than on their laptop.
When to click Stop and what cleanup happens
Stop is the partner control to Start, and it is worth treating it that way. Click it before closing or navigating away from the tab, before switching audio output devices, and any time the sound shifts from comfortable to noticeable. Clicking Stop invalidates any start that is still waiting for the browser to resume playback, stops the looping AudioBufferSourceNode, disconnects the GainNode, and closes the AudioContext. The same teardown runs if startup fails (for example, the browser refuses to create an AudioContext) or if the component unmounts, so repeated Start and Stop cycles and navigation events do not intentionally leave live audio graphs behind. If Web Audio is unavailable or the browser rejects startup, the tool shows an error and leaves no stale playing state — that is your cue to check the browser's audio settings rather than to keep clicking.
For readers who want a broader look at how digital noise works under the hood, the white noise generator explainer covers the same components from a different angle. Everything described above assumes the behavior documented in the W3C Web Audio API specification; browsers that deviate from that contract may surface the error path described in the previous paragraph instead of a working signal.