The sound

Every layer of every mode has to be able to answer why it is there. This page is that answer: how we choose the sounds and — just as carefully — where what the evidence allows us to say runs out.

Three rules before any sound

What we invoke is acoustics, not neurology. It is well established that one sound can cover another, and that some sounds bother people less than others. It is not established that a sound activates, synchronises or regulates anything. That is why you will not find binaural beats or “frequencies” here: a review of fourteen studies on binaural beats found five in favour, eight against and one mixed (Ingendoh et al., 2023).

When the evidence is thin, we say so. Several things we do are defensible aesthetic decisions rather than findings. That is fine. Selling them as science would not be.

Low volume is part of the design, not a compromise. The one study that finds a cognitive benefit in adults without attention difficulties finds it at 45 dB, and at 65 dB reports more errors and more stress (Awada et al., 2022). Use Head Reset at the lowest volume that works for you.

Focus

Distant rain in front, brown noise underneath, a very low tone behind. No notes.

The mechanism we stand behind is covering other people’s speech, not stimulating you. How much a background conversation costs you depends on how intelligible it is, and reducing that intelligibility reduces the distraction (Hongisto, 2005). That leads to a concrete decision: speech is intelligible between 250 Hz and 4 kHz, and consonants — the part that decides whether you understand a word — live between 2 and 4 kHz. Brown noise barely reaches there; rain does. So rain goes in front and brown noise underneath, not the other way around.

And there are no notes, no birds, no thunder. What interrupts working memory isn’t noise: it’s change. Steady noise does not disrupt, and making it vary is enough to start (Tremblay, Macken and Jones, 2001). For the same reason we care so much that no seam is ever audible: a joint that shows up every thirty seconds is exactly the kind of event that steals attention.

What we don’t say: that brown noise improves focus. The reference meta-analysis on noise and task performance states plainly that it found no studies of brown noise at all (JAACAP, 2024). And in people without attention difficulties, the average effect of white or pink noise in that same work is slightly negative.

Let go

Slow waves in front, wind well behind, loose notes.

The wave takes about eleven seconds, and that is not arbitrary: rhythmic phrases of around ten seconds synchronise cardiovascular responses (Bernardi et al., 2009). But the important part is different. The wave doesn’t pull you along: it sets a pace. What has a large evidence base — hundreds of studies — is breathing slowly on purpose (Laborde et al., 2022); passive entrainment, on the other hand, does not show up when you look for it directly (Juslin et al., 2026).

Hence a mixing decision: the wind sat so close to the waves that it filled the gaps between them, and the rhythm dissolved into an even wash. The wind is now well back, so a wave sounds like a wave. Breathe with it if you like, or just let it play.

What we don’t say: that it lowers your stress, improves your heart rate variability or your blood pressure. A four-week trial found no change in resting heart rate variability (Sumińska et al., 2026), and across 22 trials with more than 17,000 people, guided slow breathing did not lower blood pressure (de Freitas Gonçalves et al., 2022).

Rest

Rain from a real recording in front, a low bed underneath, a very quiet tone and, every few minutes, distant thunder.

What wakes someone is not how loud a noise is: it is the jump from quiet to that noise. In a polysomnography study, adding continuous background sound brought arousals from hospital noise back down almost to a quiet night, precisely because it shrank that jump (Stanchina et al., 2005). It is a very small study, but the mechanism is clean and it is the one we follow.

For the same reason there are no notes and no events you could count. Signals that pulse with silences in between wake people at lower intensities than continuous ones (Bruck et al., 2009): a note every thirty seconds for an hour is exactly the pattern to avoid while someone is falling asleep. The crickets that were there at first are gone too: their energy sat right in the band where the ear is most sensitive.

The rain is now a field recording, not synthesis: no matter how much we refined it, generated rain lacked the irregularity of the real thing — gusts, shifts in the wind, drops falling from leaves. It is a public-domain recording, made outdoors, with nothing that resonates. The thunder comes from the same take, and only the distant rolls that build slowly made the cut: a long, low rumble every few minutes is the background changing shape, not a jolt.

What we don’t say: that Head Reset improves your sleep. The reference systematic review rates the evidence for continuous noise as a sleep aid “very low”, and its authors advise against promoting it (Riedy et al., 2021). Nor do we say anything about slow waves or memory: the studies cited for that use real-time EEG, which no app that simply plays audio does.

Fan

A blanket of broadband air, and nothing else. No blades, no motor, no notes. It is the only soundscape with no pitch at all.

It comes from real life: plenty of people sleep with a fan on, and not always because of the heat. We didn’t want to invent anything, just reproduce that. The beat of the blades and the hum of the motor were left out for the same reason Focus has no notes: they are periodic, something the ear can follow, and what makes this mode work is that there is nothing to follow.

It isn’t white noise, and that’s deliberate. In a laboratory experiment, forty people rated how annoying they found twenty-three broadband noises with different spectra. Treble-heavy ones were as annoying as the same sound played 9 to 10 dB louder; bass-heavy ones were as annoying as the reference or less; a neutral, pink-like one landed in between (Kuusinen, Rantanen & Hongisto, 2024). White and pink noise fall on the annoying side. Fan slopes down toward the treble, on the comfortable side of that experiment, with a low body underneath that makes it sound like a machine in the room rather than something in your headphones.

Why broadband for sleep is the same as for Rest, and it applies even more cleanly here: low-level broadband sound shortened the time to fall asleep in a polysomnography trial (Messineo et al., 2017), and an even floor shrinks the jump that causes arousals (Stanchina et al., 2005). Acoustically, a fan is exactly that floor.

What we don’t say: everything we don’t say about Rest, without exception. The evidence for continuous noise as a sleep aid is of very low quality (Riedy et al., 2021).

Travel

Tuned air in front, chords, clear notes. No bass.

This mode was rebuilt from scratch after listening to it while driving. The problem was physical: a car on the motorway makes its own noise, and that noise lives exactly where the soundscape was sitting. Adding noise inside a noisy environment doesn’t stand apart from it; to hear it you have to drown it out, and at that point it no longer relaxes anyone.

A sound with pitch — a note, a chord — is audible far below the level of the noise covering it, because it concentrates its energy in a single band while noise spreads across dozens. The most forceful proof of that isn’t from a laboratory but from regulation: international environmental-noise standards add up to 6 dB of penalty to a sound purely for being tonal (ISO 1996-2). So Travel is tonal and it plays quietly. And quiet is the point: the less it covers, the better you still hear the road.

It also settles where to put it. Cabin noise concentrates between 125 and 500 Hz; the ear is most sensitive between 3 and 4 kHz (ISO 226); and in-vehicle alert standards work between 500 and 2,500 Hz (ISO 15006). Travel lives in that window, and has nothing below 90 Hz: the car already supplies the bass, and a driver is also receiving low-frequency vibration that on its own induces drowsiness within twenty minutes (Azizan et al., 2016).

What you will never find here

• Binaural beats and “healing frequencies”. There is nothing to support them.
• Bells, chimes, thunder or birds in Focus. They are events, and events steal attention.
• Efficacy figures. No “75% sleep better”, no equivalent. Numbers like that circulate widely and are almost never in the study they are attributed to.

Volume, ears, and who should not use this

The WHO-ITU safe listening standard sets 80 dB over 40 hours a week for adults and 75 dB for children. And it is worth knowing that sound is never free: at just 37 dBA of noise, less deep sleep is already measured than in silence (Myllyntausta et al., 2020). Louder is not better.

Do not use Head Reset with babies or young children. Of fourteen infant sleep machines measured, all fourteen exceeded 50 dBA at 30 cm and three exceeded 85 dBA (Hugh et al., 2014).

What we still don’t know

Three things, said plainly. First: nobody has studied this use case — a session you choose, when you want it, where you want it. All the evidence comes from laboratories or from offices where the sound was imposed. Second: believing a sound is natural changes what gets measured, regardless of how it actually sounds, so part of what these studies observe is expectation. And third: much of the literature favourable to noise for attention comes from a single research group, while the sceptical literature comes from independent ones. We weighted that when deciding.