The Loudness Wars Are Over. The Quiet Wars Have Begun.

The mastering room at Northlight Studios has a leather sofa nobody sits on and monitors worth more than most cars. Sam Okereke is ignoring both. He has laid a phone flat on the desk, speaker up, playing a rock mix at conversational volume while a kettle rattles in the corridor.

"That is the reference," he says, nodding at the handset. The monitors tell him the truth. The phone tells him what happens to the truth.

For two decades the industry had one argument about volume, and it was simple: more. Compressors were pushed until snare drums turned to cardboard, because a louder record won every comparison that mattered. Then the platforms began normalising playback, turning everything down to roughly the same perceived level, and the reward disappeared. Master at minus six and the listener still gets minus fourteen. All you keep is the damage.

Everyone agrees that war is over. Fewer people say out loud that another started in its place, fought over something harder to put a number on: how much of a record you can lose before it stops being the record.

The truce nobody signed

Okereke spent fifteen years making records louder and now refuses to, a conversion he walked us through last year. He keeps the paperwork. "In 2012 I was getting maybe thirty masters a year with a note attached saying match the reference, which always meant beat the reference," he says. "Last year, four. Two of those were reissues."

He pulls up two versions of the same band, sixteen years apart, levelled so they measure the same. The 2008 master leaves six decibels between its average level and its peaks. The 2026 one leaves eleven. Matched, the older record sounds smaller, which is the opposite of what it was built to do. "Every trick we used cost us something," he says. "Normalisation just stopped paying for the tricks."

So the file got its dynamics back. The trouble is that nobody listens to a file.

Mixing for the worst speaker in the room

Hana Brandt masters from a first-floor room above a bike shop on Istedgade in Copenhagen, with nearfields, a laptop speaker, and a car radio wired to a bench supply at head height. Nine out of ten pop sessions, she says, arrive with the same sentence in the email: make sure it works on a phone.

"Nobody writes make sure it works on a stereo," she says. "I have not read that sentence in six years."

In practice that means decisions with nothing to do with loudness. Below roughly 80 hertz a phone produces almost nothing, so the weight of a track migrates up into the band where speech lives. Vocals sit three or four decibels higher than they would have in 2010. Sub-bass gets a harmonic ghost near 200 hertz so the ear infers a note the speaker never made. Reverb tails are shortened, because a tail on a phone in a kitchen is not atmosphere, it is mud.

Brandt is unsentimental about all of this, and precise about what it costs.

We stopped arguing about how loud a record can be and started arguing about how much of it you can throw away before it stops being the record. That is the worse argument, because you can win it and be left with nothing.

None of this is new, she adds. The transistor radio was a bad speaker too. The difference is that in 1966 nobody mixed exclusively for it.

The seven-second audition

Ines Carvalho programmes three mood playlists for Halcyon from an office on Rua do Alecrim in Lisbon. Between them they carry about 6.1 million followers, which makes her one of the most consequential listeners in Europe.

"A track that loses a third of its audience before twenty seconds does not get renewed the next month," she says. "I do not decide that. The number decides and I read the number." Most of that loss, she says, happens in the first seven seconds.

Ask whether she has ever told a label to change a record and she says no, twice, and means it both times. "I have never asked anyone to shorten an intro. I have also never had to ask." The A and R notes we were shown this year do the asking for her: hook by seven seconds, no eight-bar build, cold open preferred, chorus inside the first minute.

The result is not worse music, necessarily. It is music with a different shape, and the shape is not being decided in a studio.

The bar test

The Bell Foundry seats 58 people at 14 Cotton Row in Manchester and has, by its owner's estimate, the least forgiving acoustics of any room she has worked in. "At nine on a Friday, before I put anything on, the room is 78 decibels of people," Delphine Aigner says. "The PA is 90 watts a side, under a ceiling that eats everything above head height."

Behind the till she keeps a notebook of records that survive the room. In four years it has 61 entries, and most are not obvious. "People assume a loud bar needs loud music. A loud bar needs music with something in the middle of it," she says. "Half the albums I love at home are inaudible here. They turn into weather."

That notebook is the clearest statement of the new war we found. Dynamic range has stopped being a default and become a choice with an invoice attached: a list of rooms your record will not work in. Moth Choir built an entire album out of accepting that bill. Kilo Nova's Attenuation takes the opposite bet, dropping the kick out for whole bars and trusting the reverb to hold the floor, and it survives the Bell Foundry, which is a genuinely strange fact about 2026.

Back at Northlight, Okereke's test ends the way it always ends. He picks up the phone, walks to the far end of the room where the kettle is, and stands with his back to the desk. If he can still find the snare from there, the record goes out. It is not a scientific instrument. It is, at the moment, the most honest one in the building.