Field note · lab

A drifting heartbeat, under your own music.

Tone Lab lays five kinds of tone — a binaural pair, harmonic pairs, a pulse, a slow heartbeat, a noise carrier — quietly under whatever you're already listening to. It runs live in the browser, tuned by ear over one long evening, not sold as a cure.

Pick a goal — focus, relax, meditate, sleep — and the beat ramps toward it over the length of a session, then holds until your music runs out and everything fades gently on its own. Sleep is the exception: you set its total time, and it fades out where it has reached rather than waiting on the music.

Open the lab How it works

● 01 · LiveRuns in your browser

Tone Lab

Layers: five — binaural pair, harmonic pairs, pulse, heartbeat, noise carrier
Goals: focus, relax, meditate, sleep, and more
Runs: in the browser, over your own music
Export: MP3, from the local tool

“Soft-wave harmonic pairs, and a heartbeat sliding from 62 down to 48 bpm — the recipe that actually worked, by ear.”

Open the lab

Premise

Why this exists

Binaural-beat tools tend to be clinical — a sine pair, pink noise, a timer. Generative ambient tools such as Aeon, generative.fm, and Endel are musical and layered, but carry no beat layer at all. Nothing found did both halves well.

So Tone Lab does the smaller thing: it lays those tones quietly under music you already have, instead of replacing it. Aeon's design — voices stacked over a shared clock — was the starting point; the sound underneath came from testing recipes against real tracks, by ear, one evening.

Aeon solved the beauty. This tries to solve the other twenty percent.

What it does

Three things, plainly

Five layers, stacked

A binaural pair, harmonic pairs, a pulse, a drifting heartbeat, and a noise carrier — turn any of them on under your track. Nothing installed, nothing uploaded.

Goals that move the beat

Focus, relax, meditate, and a few others each ramp the beat over the arc of a session, hold at the target, then return gently to 10 Hz when the music ends. Sleep runs on its own clock instead — a total session time you choose, ending in a two-minute fade at whatever beat it has reached. Both are comfort choices about how a session should finish, not findings about what a beat does to you.

Not a claim

How much evidence there is depends on the protocol and the outcome measured — some studies find effects, others don’t, and EEG entrainment itself is unsettled. Nothing ties binaural beats specifically to deeper meditation. These presets are not clinically validated, and this is not medical.

The recipe

What each goal is for

A goal sets the target beat band and how a session gets there. It says nothing about timbre — any recipe of layers can carry any goal, though the recipes are not interchangeable as stimuli: the brainaural-ish one has no binaural pair in it at all. Choosing a goal never changes your sound settings. Evidence depends on the full protocol, including frequency, and on the outcome measured. A band name alone does not establish a benefit, so the table describes each preset without grading its effectiveness.

Goal ↔ beat band ↔ what the preset is
Goal	Target band	What it is
Focus	10 → 14 Hz, low beta / SMR	A listening intent. The band is borrowed from SMR neurofeedback — a different intervention that passive audio has not been shown to reproduce.
Relax	12 → 10 Hz, alpha	The long-standing relaxation setting. Some anxiety and relaxation findings are positive, others are not.
Meditate	10 → 6 Hz, alpha → theta	The main use case. A recent paper reporting two experiments found better self-rated calmness and focus at 6 Hz — self-report, not meditation depth.
Deep meditate	10 → 6 → 4 Hz, theta	A slower, longer descent — a different shape, not a deeper state. Nothing shows it goes further than 6 Hz, and no minimum time is required.
Sleep	10 → 6 → 3 → 1.5 Hz, alpha → theta → delta	Experimental wind-down. Results vary by protocol and outcome, and there is no established 1.5 Hz destination. The only goal with its own total session time, fading out where it has reached; the length is a comfort setting, not a dose.
Body scan	8 → 4 Hz, delta–theta	Theta-beat trials suggest a possible adjunct for pain, on low-certainty evidence. Not healing, and not a validation of this arc.
Energize	10 → 18 Hz, beta	A listening intent. Beta beats and attention are mixed; one careful pair of experiments found evidence against a 16 Hz benefit.
Creativity	10 → 7.5 Hz, theta/alpha border	Speculative. Reported creativity effects varied by task and by person, and nothing singles out 7.5 Hz.
Gamma	40 Hz, held	Experimental, and above the ~30 Hz range where beats are usually reported as clearly perceptible. The 40 Hz Alzheimer’s work is light and sound clicks, not binaural beats.

By ear — one listener, one evening, a preference and not a finding — one stack won: soft-wave harmonic pairs under a plain binaural pair, plus a heartbeat panned slowly and drifting from a resting 62 bpm down to 48 over the session. A bare sine pair alone, without any of that, was rejected outright — the lab notes call it, simply, “sucks.”

Notes

Depth, if you want it

What the research supports ↓

It depends on the protocol and on the outcome. A 2019 meta-analysis ↗ of 22 studies (35 effect sizes) reported a pooled g of about 0.45 across memory, attention, anxiety and analgesia — a standardised effect size, not a percentage, and not this product's effect. Exposure timing and duration moderated the results. A 2023 review of EEG studies ↗ found entrainment itself unsettled: five studies in line with it, eight contradicting it, one mixed, and "no clear trend discernible either in favor of the brainwave entrainment hypothesis or against it." An earlier version of this page called that review "encouraging but mixed" — that was too favourable, and it is corrected here.

For the main use case, a 2026 paper reporting two experiments ↗ (101 and 118 participants, five-minute exposures on a 250 Hz carrier) found 6 Hz beats significantly increased self-rated calmness and focus, the second experiment against pink-noise and silence controls. Those are self-report scales: the authors note that without EEG or fMRI, "conclusions regarding neural entrainment remain speculative." Pink noise alone moved calmness too. Nothing there validates the ramps or the layered stack used here.

For sleep, a double-blind trial ↗ in subclinical insomnia found that adding theta beats to music for 30 minutes before bed did not improve insomnia severity more than music alone — while a 2022 Cochrane review ↗ of 13 trials and 1,007 participants found moderate-certainty evidence that music improves subjective sleep quality, with its other sleep outcomes less certain. The music may be doing much of the work. That is a useful result, not a disappointing one.

Two specifics behind the table. A 2026 review of theta-beat trials ↗ pooled four pain trials (173 participants) at SMD −0.53 on low-certainty evidence, describing beats as a possible adjunct for pain modulation — an adjunct, not healing, and not a validation of the Body scan arc. And two experiments on 16 Hz beats and sustained attention ↗ reported evidence against a benefit rather than merely failing to find one.

Some things this page used to say and no longer does. There is no established best carrier range: 200–400 Hz was never a research sweet spot. G3 ≈ 196 Hz here is a musical choice; 250 Hz is useful mainly for comparing with the studies that used it, which chose it for comparability rather than for effect. There is no nine-minute threshold and no rule that longer is always better. The return to 10 Hz at the end of a session is a design choice for a gentle landing, not a physiological requirement. Beats are generally reported as clearly perceptible up to around 30 Hz, which is an approximate boundary rather than a wall.

The layers are different methods and do not share each other's evidence. A binaural pair is two tones, one per ear, with the beat produced by the listener — it needs headphones, because on speakers the tones mix in the air. The pulse and the noise carrier are amplitude modulation: a single carrier varied in level at Δ. They play perfectly well on speakers, which is exactly why binaural research does not transfer to them. Setting stereo width to 0 mixes both tones into both ears, so that is not a binaural pair either.

None of this shows that binaural beats deepen meditation, and the complete presets here have not been clinically validated. Tone Lab is experimental audio for relaxation, meditation practice and personal listening comparisons. It isn't medical.

Implementation notes ↓

Everything here runs in plain Web Audio — no build step, no server, no account. The tones are computed live in your browser from the settings in the page's own URL, so a link is also a saved mix.

Locally, the command binaural builds a small Python virtual environment, starts a static server, and opens the lab. lab/render.py — numpy plus ffmpeg — renders the current mix to an MP3 in exports/, around 40× faster than real time. That export step only exists on the local machine: the hosted page runs the identical tones live in the browser, with no export button, because there's no Python behind it to run one.

The lab deploys as one static HTML file, plus a generated list of the tracks on offer, to loomisresearch.com/labs/binaural/.