How We FFT-Verify Every Session

A noise track makes a claim before it plays a single second. The label on it, white, pink, brown, or green, is a promise about the shape of the sound: how its energy is distributed across the frequencies you can hear. That promise is not a feeling. It is a number, and the number can be checked.

Most channels never check it. They upload a file, type a color into the title, and move on. We do the opposite. Every session we publish is built from a mathematical definition, then measured against that definition before it ever leaves the lab. This piece is about that process: what the label really claims, what the research behind these sounds can and cannot support, and how you can hold any track, ours included, to its word.

What the claim actually is

Noise colors are not moods. They are spectral slopes, a precise statement about how loud each frequency is relative to the others. Plot power against frequency on a log-log graph and each color traces a specific line.

White noise is flat: equal energy at every frequency, a slope of 0 decibels per octave. Pink noise tilts downward at roughly 3 decibels per octave, which is why it sounds fuller and less harsh than white. Brown noise is steeper still, falling off at about 6 decibels per octave, so the highs are rolled almost all the way down and what remains is that deep, soft rumble. Green noise is the odd one out: instead of a single downward slope, it concentrates energy in a mid-band emphasis near 500 Hz, roughly where human hearing is most sensitive.

Each of those is a testable claim. A file is either following its color’s line or it is not. When a track sold as pink noise actually measures nearly flat, it is not pink, no matter what the title says, and it will sound and behave differently from the real thing. The label is a specification, and a specification can pass or fail.

What the science actually shows

It is worth being honest about how much weight these sounds can carry, because that is exactly why the label has to be accurate.

The strongest, most widely accepted benefit of any steady sound is masking: a constant, even sound covers the sudden intrusions that break concentration or wake a light sleeper, a door, a notification, a voice down the hall (Capezuti et al., 2022). That mechanism is well understood and not controversial.

Beyond masking, the evidence thins quickly. A 38-study systematic review of continuous noise as a sleep aid concluded that the overall quality of that evidence is low, with small samples and mixed designs across the literature (Riedy et al., 2021). For attention, the picture is similarly modest: white or pink noise produced a small positive effect in young people with elevated attention difficulties (Nigg et al., 2024). These are real but narrow findings, tied to specific colors in specific groups.

That narrowness is the whole point of measuring. If the one study that found an effect used pink noise, and the file you are playing is actually flat, you are not getting the sound that was tested. The research is only as meaningful as the accuracy of the label it is attached to.

What we measured

We do not use field recordings, and we do not trust a color name we cannot verify. Each session starts from the mathematical definition of its color, generated in Python, then checked with a Fast Fourier Transform the way an instrument would check it.

Here is the receipt for the four noise colors, measured: white sits at 0 decibels per octave, flat across the audible band. Pink falls at -3 decibels per octave. Brown falls at -6 decibels per octave. Green shows its mid-band emphasis near 500 Hz rather than a straight slope. Every one of those numbers comes from the FFT of the actual rendered file, not from the label we hoped to earn. When the measured slope does not match the definition, the render fails and goes back for another pass.

We publish that slope in the description of every session, because it is the one part of the claim that is not a matter of taste. You should not have to trust that a track is what it says. You should be able to read the number, and then check it yourself. The full walkthrough of this pipeline, with the real spectra on screen, is the subject of our Behind the Lab session.

The honest limits

Measurement is powerful and also strictly bounded. Verifying the slope proves what the sound is. It says nothing about what the sound will do for you, and we will not pretend otherwise.

There is, for example, no direct peer-reviewed trial of brown noise for attention or ADHD; the small attention effect in the literature was found with white or pink noise, in youth, not with brown (Nigg et al., 2024). And more is not always better: a 2026 study found that continuous pink noise played through the night reduced REM sleep, a reminder that a technically correct sound can still have trade-offs worth naming (Basner et al., 2026). A verified slope is a floor, the minimum honesty we owe you, not a ceiling of benefit.

So the measurement settles one question completely and leaves the harder question open. What the sound is, we can prove. Whether it helps you, on your task, on your night, is something only your own experience can answer.

Verify it yourself

You do not have to take our word for any of this, and you should not take anyone’s. The tool is free and the test takes less time than reading this section.

Drop any noise track into a free spectrum analyser and look at the shape. White should read flat. Pink should tilt down at about 3 decibels per octave. Brown should fall in a clean straight line at about 6 decibels per octave on a log-log plot. Green should bulge in the mid-band rather than slope. A great many files sold under these names measure nothing like their label, and once you have seen the plot, you cannot unsee the mismatch.

The claim and the measurement should agree. On our sessions, and on anyone else’s, when they do not, you have learned something useful, and it took thirty seconds to learn it.

Not medical advice. These sessions support relaxation and general wellness, and are not a treatment for any condition.