We generate sleep sounds instead of looping them

Most sleep apps ship you a thirty second recording of rain and play it eight hundred times. It works until the night you notice the join, and after that it is the only thing you can hear. Building non looping sleep sounds is the fix, and on a phone it turns out to be cheaper than shipping the audio would have been.

Your ear is a pattern detector that does not switch off

Hearing is the sense that stays on watch. You can shut your eyes, you cannot shut your ears, and the auditory system spends the night doing one job: filtering out what is predictable and flagging what is not. That is why a fan disappears within a minute and a single unexpected creak on the stairs will bring you all the way up.

A loop is a gift to that system. Somewhere in the file there is a splice, and even a well made one leaves a fingerprint: a tiny discontinuity in the noise floor, a bird that starts the same phrase at the same moment, a gust that always arrives four seconds after the last one. You will not consciously hear it on the first pass. Your auditory system will, because finding repeats is the entire job, and once it has locked onto the period it will keep announcing it.

Then you notice it consciously, and it is over. A seam is a small novel event arriving on a schedule, and a small novel event is precisely the thing that pulls you back toward wake. The sound you chose to help you fall asleep is now a metronome counting how long you have been failing to.

What non looping sleep sounds actually are

Tock ships 63 fall asleep sounds. Twenty two of them are bundled recordings, because some things cannot be synthesised convincingly and should not be. A real thunderstorm has a texture that no model of a thunderstorm has, and pretending otherwise makes a worse product than admitting it.

The other forty one are generated on the device, sample by sample, while you listen. There is no file. There is a description of how the sound behaves, and the phone builds the audio in real time from that description, which means the waveform coming out of your speaker tonight has never existed before and will not exist again. There is no seam to find because there is no join anywhere in it. Nothing downloads. Nothing repeats. Two nights of the same rain are the same rain and different weather.

This is the natural way to build the noise colours in particular. White, pink and brown noise are defined by how energy is distributed across the frequency spectrum, which is a rule rather than a recording. Generating them from the rule is both more honest and better sounding than looping someone’s capture of a rule.

The distance control, and why it exists

Forty two of the sounds carry a Far, Near or Close setting. It is not a volume slider with better marketing.

Distance in the real world is a filter, not an amplitude. A storm two valleys away has had its high frequencies eaten by the air between you and it, so it arrives soft and rounded with the crack taken off. The same storm overhead keeps the transients, and the transients are the part that startles you. Rain on a window two rooms away and rain on the glass beside your head are different sounds, not the same sound at different loudness.

So Far, Near and Close change the spectrum and the reverberation rather than just the level. In practice most people who cannot settle to a sound try Far before they conclude they dislike it, and quite often that was the only thing wrong.

What a sound can and cannot do for falling asleep

Worth being straight about this. Sound does two useful things and neither of them is sedation.

The first is masking. A steady broadband texture raises the noise floor of the room so the sudden events that would have crossed your waking threshold, a car door, a neighbour, a pipe, no longer stand out enough to register. That reduces the chance of a nocturnal awakening you would otherwise have had, which is the reason sound helps as much in the middle of the night as at the start of it.

The second is occupying attention. A lot of long sleep latency is not physical, it is a mind that will not stop producing sentences. Something mildly interesting and completely undemanding to listen to gives that machinery a low grade task, and low grade tasks are much easier to fall asleep out of than silence is.

What sound will not do is generate sleep onset on its own. If you got into bed three hours before your body clock wants you there, the rain is not going to win that argument, and no sound fixes a schedule. That part lives in why you wake up at 3am if the problem is the middle of the night, and in how to wake up easier if the real trouble is at the other end.

The same argument applies to alarms

Wake tones get the opposite treatment on purpose. There are eighteen of them and you can use your own audio, and none of them are trying to be unpredictable, because a wake tone should be immediately recognisable as the thing that means get up. Predictability is a bug at midnight and a feature at seven.

You can hear the whole set on the sounds page before you install anything. The test worth running is the boring one: put a sound on, leave it running for forty minutes while you read, and see whether you start hearing the shape of it. With a loop you eventually will. That is the entire reason we stopped shipping them.