The science, honestly

Brainwavest is built on the frequency-following response — the observation that periodic sound evokes time-locked activity in the auditory system. That much is solid. What that activity then does to attention, sleep or mood is a genuinely open question. This page describes every mechanism in the engine in full, and then draws the line between what is measured and what is hoped.

The signal chain

Nothing is streamed and nothing is stored. Every sample of sound is calculated on your device as you listen — 48,000 numbers a second, for each ear — from the session’s published specification. That calculation runs inside an audio worklet: a small program the browser gives its own dedicated audio thread, which is why a session does not stutter when the rest of the page is busy and why it keeps playing with the tab in the background.

Two gain stages, deliberately separated. The session stage carries the fades; the master stage carries your volume and nothing else — so changing the level mid-session cannot disturb a scheduled fade-out, and vice versa. Every fade is scheduled on the AudioContext clock rather than on animation frames, which is why a session still ends on time with the screen off and the tab in the background.

The second stage places the sound in space using an HRTF — a head-related transfer function, the filtering your own head, shoulders and outer ears apply to sound arriving from a particular direction. Simulating it is what lets headphones put a sound behind or beside you rather than just louder in one ear. The last stage is a limiter, a ceiling that refuses to let the signal exceed −1.5 dB, so nothing can clip and distort however the parts combine.

Five synthesis modes

The engine can carry the same beat rate five different ways, and they are not interchangeable. The distinction that governs everything else is whether the rhythm is physically present in the air or assembled inside your head.

Isochronicmonoworks on speakers: One carrier, gated by the envelope at the beat rate. The pulse is a real change in loudness, so a speaker reproduces it exactly.
Binauralstereoneeds headphones: Each ear gets a steady tone; the two differ by the beat. Neither channel contains a pulse — the beat is the interaural difference.
Monauralmonoworks on speakers: Both tones are summed before they leave the bus, so they interfere acoustically. Smoother than isochronic, and it survives a speaker.
Phase-amplitudemonoworks on speakers: A fast gamma rhythm whose loudness is itself governed by a slow one — the two envelopes multiplied rather than mixed.
Poly-rhythmicstereoneeds headphones: Each ear gets its own carrier and its own beat, related by the golden ratio rather than driven in lockstep.

Every row carries the same beat rate; only the delivery changes. In the top three the trace visibly swells and contracts — that pulsing is the beat, present in the air, which is why a speaker reproduces it. In the binaural row both channels are perfectly steady: neither contains a pulse anywhere, and the beat exists only after the two ears are compared, which is why it needs headphones and why it vanishes the moment the channels are allowed to mix in the air. Monaural is the instructive middle case — it also uses two tones, but sums them before they leave the bus, so the interference happens in the signal rather than in you.

Of 777 sessions, 610 ship as isochronic and 52 as phase-amplitude coupled. Both put the beat in the signal itself — a deliberate choice rather than a limitation, since an amplitude-modulated tone drives a measurable steady-state cortical response reliably enough that the 40 Hz case has been a standard laboratory measurement since 1981². A binaural beat, which never exists acoustically, produces a weaker and less consistent response. Binaural is offered on 688 sessions and monaural on 754, switchable mid-session without restarting it.

Where the signal goes

Sound becomes nerve activity at the cochlea and ascends through a chain of relays. One of them matters more than the rest here.

The superior olivary complex is the first place input from both ears meets. A binaural beat cannot exist before that point, which is precisely why it depends on interaural timing — and why it fades out when the carrier climbs past roughly a kilohertz.

The superior olivary complex is the first point at which input from the two ears is compared. A binaural beat is constructed there rather than heard, which is why it needs stereo headphones to exist at all, and why recordings from single neurons there find them phase-locking to the beat rate — firing at a consistent point in each cycle — rather than to either tone on its own.

Why some sessions refuse binaural

Binaural beating depends on the auditory system resolving a timing difference between the ears, and that ability has limits which were measured seventy-five years ago. Perception fails above roughly a 1 kHz carrier and needs the two tones to differ by only a few tens of hertz¹; it is clearest with carriers in the 400–500 Hz region.

The engine offers a mode only where it can physically work. Above roughly 40 Hz the auditory system can no longer resolve the interaural timing difference a binaural beat depends on, so those sessions are isochronic or phase-amplitude coupled and the option is simply not shown.

So the engine offers a mode only where it can physically work. Every carrier in the catalog sits between 80 Hz and 963 Hz, inside the classical limit. The 89 sessions that cannot be switched to binaural are the ones targeting beats above 40 Hz, where the effect stops forming. Monaural has a second guard of its own: the two tones must differ by no more than 12% of the carrier, above which they stop fusing into a beat and separate into a chord.

The envelope

In every mode except binaural and monaural, the beat is produced by shaping the loudness of a tone. The envelope is the shape of one pulse — and it is the difference between a session that pulses and one that ticks.

The envelope is the shape of one pulse: how the tone rises, how long it holds and how it falls away. Every shape starts and ends at exactly zero, so the sound can never step-discontinue into the silent half of the cycle — that discontinuity is what a click is. Almost every session runs a 50% duty cycle, so the pulse and the gap are equal and what changes between them is the edge — but duty is a real parameter, and the Frontier collection uses it: a 0.1 duty is sparse and metronomic, a 0.9 duty is very nearly continuous.

Two parameters ride alongside it. The duty cycle is the fraction of each pulse cycle that actually contains sound; every session here runs at 50%, 38%, 20%, 80%, 25%, 85%, 10%, 33%, 66%, 75%, 90%, 15%. At 50% the pulse and gap are equal; lower values create a shorter pulse, while higher values keep the pulse present for more of each cycle. Depth is how far the loudness falls between pulses, ranging from 0.5 to 0.98 across the catalog — at 0.5 the tone merely dips, at 0.98 it very nearly disappears and returns. A shallow depth is gentler and less intrusive; a deep one is a stronger stimulus.

The noise bed

Underneath the tone sits a noise field. It does two jobs: it masks the room you are actually in, and it gives the ear something broadband to rest on so the carrier does not become the only thing in the world.

Four generators, plus a spectral tilt applied on top — a one-pole low-pass with self-calibrating makeup gain, so darkening a bed does not also quieten it. That tilt is what separates “Brownian” from “Deep Brownian” from “Sub-Bass Pink”: without it those would be one signal wearing three names. Level and color stay independent controls, which is why the catalog can name 19 distinct beds from four sources.

Four generators and a spectral tilt produce the 17 distinct bed names in the catalog. The tilt is a one-pole low-pass — the most economical way to remove high frequencies, nudging each output sample a fixed fraction toward the input — with self-calibrating makeup gain, which puts back the loudness that filtering takes away.

That second part matters more than it sounds. How much level a low-pass costs depends on the noise’s own slope, so a textbook correction that is exact for white noise overshoots brown by roughly eight times and clips. Measuring the loss directly instead of assuming a spectrum is the difference between a dark bed and a distorted one.

Carrier drift

An optional treatment, switchable from the player on any isochronic or phase-amplitude session: instead of one fixed carrier, a Shepard-Risset glissando.

A perpetual glissando. Partials an octave apart sweep together through a fixed four-octave window centered on the carrier — a 216 Hz session sweeps roughly 54–864 Hz. Each partial fades in at one end of the band and out at the other, so the pitch seems to fall indefinitely without ever leaving a comfortable range. The beat rate is untouched; only the timbre moves. It is offered on isochronic and phase-amplitude sessions only: binaural, monaural and poly-rhythmic all encode the beat in an exact relationship between two carriers, and replacing one of them with a moving stack of octaves would destroy the thing they carry.

Jitter, and why nothing here is perfectly steady

A perfectly fixed carrier is a problem. Nervous systems stop responding to anything perfectly unchanging — habituation, the same reason you stop noticing a fan you have been sitting next to. For an unwavering tone that takes roughly fifteen to twenty minutes, and the tracking response falls away with it, which is awkward for a catalog whose longest session runs 95 minutes.

So every session carries a stochastic micro-instability — stochastic meaning governed by chance rather than by a fixed pattern, so it wanders unpredictably instead of wobbling regularly. Two independent random walks, one on the carrier and one on the beat, drift against each other rather than in lockstep. The deviation is about 0.1%: a 200 Hz carrier wobbles by roughly 0.2 Hz and a 10 Hz beat by about 0.01 Hz. That is far below the ear’s pitch-discrimination threshold, so you cannot hear it, and far below anything that would disturb the target rate. The tail is bounded at 2.5× nominal, because a rare large excursion would be an audible pitch slide rather than the micro-wobble this is meant to be.

This is the most speculative mechanism on the page and we would rather say so. The habituation it addresses is well documented; that this particular counter-measure meaningfully prevents it is a design inference, not a measured result.

The spatial orbit

The whole field is placed in a virtual position around your head using an HRTF panner, and that position moves — slowly. Orbit rates across the catalog run from 0.005 Hz to 0.08 Hz, which is one full revolution every 13 to 200 seconds. At the slow end it is barely a drift.

It is there for texture rather than for entrainment, and we make no claim for it beyond that: a static field becomes predictable, and a moving one keeps a long session from settling into wallpaper. On speakers the HRTF cues largely collapse, which costs the sense of position but nothing of the beat in the three mono modes.

Two sessions, in full

Overnight Delta Drift

A 108 Hz carrier, amplitude-modulated from 10 Hz down to 0.5 Hz across 39 minutes, on a sigmoidal envelope at 0.95 depth. Underneath, a brownian bed. The field orbits at 0.01 Hz. Because the pulse is a real change in loudness it survives a speaker, and because the descent is gradual rather than a step there is no point at which the sound audibly changes gear.

Flow State Gamma-Alpha Bridge

Phase-amplitude coupling: a 40 Hz gamma rhythm whose loudness moves under a slower 10 Hz to 40 Hz control rhythm, the two envelopes multiplied rather than mixed, on a 360 Hz carrier with a violet bed. The structure is borrowed from a real feature of cortical activity, where the phase of a slow oscillation modulates the amplitude of a faster one. We are reproducing that shape in sound. We are not claiming to reproduce it in you.

170 sessions hold a constant beat; the other 607 ramp between two rates, and 48 follow a three-point staged curve that settles at a middle waypoint on the way. Every session page draws its own curve before you press play.

What the evidence actually supports

There are three separate claims here with very different amounts of support behind them. Running them together is the central dishonesty of this category.

Well established

Periodic sound evokes time-locked electrical activity in the auditory system. The 40 Hz auditory steady-state response is reliable enough to be used as a clinical and pharmacological measurement².

Genuinely contested

That binaural beats entrain broader brain oscillations. A 2023 systematic review of 15 EEG studies found 5 supporting entrainment, 8 contradicting it and 1 mixed, with heterogeneity high enough to limit comparison between studies⁴.

Suggestive, not settled

Behavioral outcomes. A 2019 meta-analysis across 22 studies and 35 effect sizes reported a significant pooled effect on memory, attention, anxiety and pain (Hedges’ g = 0.45), varying with frequency, exposure length and timing³. That is a real result from a small, heterogeneous literature — not a settled mechanism.

One more caveat we would rather state than bury: most of that research studies binaural beats specifically, while most of what we ship is isochronic. The physical case for amplitude modulation driving a steady-state response is stronger; the behavioral literature on isochronic stimulation specifically is thinner than the binaural one⁵. We think the trade is the right way round. We are not going to pretend it is settled.

What we do not claim

That any session treats, cures, prevents or diagnoses any condition. It does not, and Brainwavest is not a medical device.
That a stated frequency puts your brain at that frequency. A session named for delta is built around a delta-rate stimulus; that is a property of the audio, not a promise about your EEG.
That the spatial orbit, the carrier drift or the jitter change any outcome. They are there to keep a long session from becoming predictable, and we describe them as craft rather than as mechanism.
That effects are uniform. The response varies between people and between days, and a meaningful share of any felt benefit is likely expectation — which is true of most things that help people relax.
That we have run trials. We have not. We publish the complete signal specification for every session so that what we are actually doing is inspectable, rather than asking you to trust a claim.

If you have epilepsy or a seizure history, photosensitivity, are pregnant, or have a heart or neurological condition, talk to a clinician before using rhythmic audio stimulation. The fuller version of that is on the safety page.

Every term, in plain words

Each of these is expanded where it first appears above. They are gathered here so nothing on this page depends on already knowing the vocabulary.

DSPdigital signal processing: Arithmetic performed on the sound itself. Rather than playing a recording, the engine calculates each individual sample — 48,000 numbers per second, per ear — from the session’s specification.
Audio workletthe sound generator: A small program the browser runs on its own dedicated audio thread. That isolation is why a session does not stutter when the rest of the page is busy, and why it keeps running with the tab in the background.
Carrierthe pitch you hear: The underlying tone. On its own it is a steady note; everything else on this page is a way of shaping it. Carriers in this catalog run from 80 Hz to 963 Hz.
Beatthe pulse rate: How many times per second the sound pulses, or — in binaural mode — the difference in hertz between the two ears. This is the number a session is named for, not the pitch.
Envelopethe shape of one pulse: How a single pulse rises, holds and falls. A square envelope clicks on and off; a bell-shaped one swells. Same beat rate, entirely different character.
Duty cyclesound versus silence: The fraction of each pulse cycle that contains sound. At 50%, used throughout this catalog, the pulse and the gap are equal.
Depthhow far the volume dips: How much quieter the sound gets between pulses. At low depth the tone merely wavers; at high depth it almost disappears and returns.
HRTFhead-related transfer function: The filtering your own head, shoulders and outer ears apply to sound arriving from a particular direction. Simulating it lets headphones place a sound behind or beside you rather than simply louder in one ear.
Interauralbetween the two ears: Anything comparing left against right — timing, level, or phase. A binaural beat is an interaural effect, which is why it cannot survive the channels being mixed in the air.
Frequency-following responsethe brain tracking a rhythm: Electrical activity in the auditory system that follows the timing of a repeating sound. It is measurable and well documented; what it means for how you feel is the open question this page is careful about.
Steady-state responsea sustained tracking signal: The same idea measured over a continuous stimulus rather than single clicks. The 40 Hz case is reliable enough to be used as a clinical measurement.
Partialone tone within a sound: A single frequency component. The Shepard drift stacks four partials an octave apart and slides them together, which is what produces a pitch that seems to fall forever.
Octavea doubling of frequency: Two notes an octave apart differ by a factor of two — 220 Hz and 440 Hz. Musically they sound like the same note at a different height.
Glissandoa continuous slide in pitch: A smooth sweep rather than a step between notes.
One-pole low-passa simple treble filter: The most economical way to remove high frequencies: each output sample is nudged a fixed fraction toward the input. It is what turns a bright noise bed into a dark one.
Makeup gainvolume put back after filtering: Removing frequencies also removes loudness. Makeup gain restores the level so that changing a bed’s color does not also change how loud it is.
Limitera ceiling on loudness: A last stage that refuses to let the signal exceed a set level, so nothing can clip and distort no matter how the parts combine.
Stochasticrandomly varying: Governed by chance rather than a fixed pattern. The carrier jitter is stochastic: it wanders unpredictably instead of wobbling regularly, which is the point.
Habituationtuning something out: The nervous system’s tendency to stop responding to anything perfectly unchanging. It is why an unwavering tone fades from awareness after fifteen or twenty minutes.
Phase-lockingfiring in time with a rhythm: Neurons discharging at a consistent point in a repeating cycle rather than at random times within it.
Hedges’ ga measure of effect size: How large a difference is, in standard deviations, corrected for small samples. Roughly: 0.2 is small, 0.5 moderate, 0.8 large. It says nothing about whether a result replicates.

References

Every term used here — carrier, beat, ramp, envelope, the EEG bands, the noise colors — is defined in plain language in the questions and glossary.