The Sound You Stop Hearing

A complaint that is also the design

The note arrives regularly and it is completely fair: the ambience is too low.

What makes it hard to answer is that the person is right about the first minute and wrong about the fortieth, and there is no level that is correct for both. A mix has a time axis. Most mixing advice quietly assumes the listener is arriving fresh and attentive and will stay that way, which is true for a song and false for anything meant to run for eight hours next to a person doing something else.

So we mix for a listener who is forty minutes in. That decision costs us the opening. We know it costs us the opening. We've never found a version where it doesn't.

Habituation, and why it ruins everything

The mechanism underneath is unremarkable and universal: a constant, unchanging stimulus stops being reported.

You are not currently feeling your socks. You were not hearing your fridge until this sentence. Nothing about the signal changed — the reporting changed, because a stimulus that carries no new information gets pushed below the threshold where it's worth interrupting anyone about. Ears do this faster and more completely than most senses, and they do it to exactly the sounds we make.

Which means a level is not a fixed property of a mix. It's a curve. The same track at the same volume is loud at minute one, correct at minute eight, and slightly thin at minute ninety, and if you set it by ear at the start you have set it for the least important minute of the whole episode.

An ambient track succeeds by disappearing. Anything still asking for attention at minute forty has failed, however beautiful it is.

Why the first sixty seconds are deliberately thin

Given all that, the opening could be handled two ways, and we've shipped both.

The first way is to start at the level that will feel right later — which means starting loud, because it has to survive habituation. This gives an excellent first impression. It also produces a distinct moment, somewhere around minute fifteen, where the rain is doing more than the room needs and the listener notices the mix as a mix. Several people described that version as pushy without being able to say why.

The second way, which is what we ship now, is to start slightly under and let attention come down to meet it. The first minute feels reserved, almost tentative. Then it stops feeling like anything, which is where it stays for the remaining seven hours and fifty-nine minutes.

The picture side of the workshop reached the same conclusion by an entirely different road, which is one of the reasons we trust it. The four-minute cat is about refusing to reward vigilance with events — never giving the eye a reason to keep watch. This is the same refusal, aimed at the ear.

What narrowing the range costs

There's a trade here and it's real, so it's worth putting on the table rather than in a footnote.

To make something sit under attention for hours, you narrow its dynamic range: the loud parts come down, the quiet parts come up, and the distance between them shrinks. It's the difference between weather and a recording of weather. Real rain has an enormous range, with lulls that nearly vanish and gusts that are startling.

Narrow that range and you get consistency, which is what a long evening needs. You also lose something, and we should name it honestly: you lose the sense that anything is at stake outside. A storm mixed this way is no longer dramatic. It cannot build, it cannot break, and it cannot surprise you, because every one of those requires range we deliberately removed.

That is a genuine loss and not everyone wants to make the trade. If what you want from rain is the drama of it, our version is going to sound flat to you, and the honest answer is that you should go and stand at a real window, where the range is free and enormous.

The seam, where repetition becomes audible

Habituation has a failure mode, and it's the thing we spend most of our production time defending against.

The ear stops reporting a constant sound. It does not stop reporting a pattern. If material repeats on a fixed cycle, something eventually catches the cycle — usually not consciously, usually reported afterwards as the track being restless or oddly tiring — and once caught, it cannot be un-caught for the rest of that listening session.

So the layers are built at deliberately mismatched lengths, so that their combination doesn't come around again for hours, and the joins are placed under moving material rather than in the open. It's a slow, unglamorous, entirely invisible discipline, and the invisible seam is the piece that goes into how it's actually done.

The short version: the quieter and more constant a bed is, the more brutally any pattern in it stands out. Mixing low buys you calm and costs you all your hiding places.

The ceiling, for the person who is already asleep

The last constraint is the one that overrides the others, and it isn't a mixing question at all.

A meaningful number of people fall asleep with these on. That is entirely their business, and we've never once pitched an episode as a sleep aid — we make ambience, we're not qualified to make claims about anybody's night, and we don't. But it does mean the mix has one hard rule: nothing in the last six hours may be loud enough to wake somebody, no matter how good it would sound at minute one.

That's why our thunder stays two hills away. Why no bird ever arrives close to the microphone. Why the fire settles rather than cracks. Every one of those choices makes the episode less interesting to sit and listen to attentively, and every one of them is the right call for a room at three in the morning with somebody in it.

No correct level exists

Which brings us to the limit of everything above.

This mix falls apart on a bad speaker. A phone lying face up on a table loses the low end entirely, and the low end is where most of our warmth lives — what's left is the thin top of a rain track, and it sounds exactly as unimpressive as that description suggests. On those speakers the people who say it's too quiet are not mistaken. They're accurately describing what reached them.

There is no universal right level. There's a room, a speaker, a distance, a listener, and a time of night, and we can't see any of the five. What we have instead is an assumption — somebody sitting a few metres away, doing something else, for a long time — and every number in the mix is derived from it. When the assumption fits, the episode disappears into your evening and you never think about any of this. When it doesn't, please just turn it up. The volume knob is more informed than we are.

The evening is warm here.