Here is what I was sure of a week ago. Everybody went home in March 2020. Half the records of the next two years were mixed in bedrooms with a laptop, a pair of monitors on a desk, and nothing on the walls. A room like that is bright. It flutters. It throws the top end back at you from the window behind the screen. My guess was that all of that high-end garbage landed on the records, and that if I pulled the measurements for 2020 through 2022 I would see a bump in brightness you could point at.

So I pulled them. We have measurements on 120,392 tracks with a release date between 2018 and 2024, the same analysis that runs under the Sonic Analyzer and the same table the tonal balance by genre study came from. Brightness there means spectral centroid, the energy-weighted average frequency of the track, in Hz. I split the years by quarter, by genre, and by how big the release ended up, because a record with ten million plays was not mixed in a bedroom and a record with four thousand plays very likely was.

I was wrong about the top end. I was right about the room.

Brightness did not move

Medians, so one loud outlier cannot drag a year. The pandemic years are not brighter. If anything the line is drifting down, about 50 Hz across seven years, which is less than the difference between two takes of the same song.

I checked whether a change in what got released was hiding a change in how it sounded. Dance and electronic releases, the brightest lanes, got darker from 2019 to 2022 (dance from a 3,294 Hz median to 2,963), so a thinner dance year could pull the average down and mask a bedroom-brightness bump everywhere else. It does not. Measured against each genre's own median, the 2020 and 2021 quarters sit within 30 Hz of zero. The top octave's share of energy moved by about one percent in either direction, and that is the noise floor of this kind of measurement.

So no. The bright-room theory does not survive the numbers. People in bright rooms did not make bright records.

What moved instead

Two things, and they are the two things an untreated room is actually bad at.

The low mids went up. The 250 to 500 Hz share of energy climbed from 8.8 percent in 2019 to 9.1 percent in 2021 and stayed there. That sounds small until you count records instead of averaging them. Take the 2019 top quartile for low-mid energy as the line for "muddy", and ask how many releases cross it each year:

Both groups jumped in 2020. Everybody was home, including the people who make the big records. Then the big records drifted back toward the 2019 line as studios reopened. The small ones never did. Thirty percent of small releases in 2024 carry more low-mid weight than three quarters of 2019 did.

The presence region went down. Count the records that fall below the 2019 bottom quartile for 2 to 4 kHz energy, the region where a vocal bites and a snare cracks:

Same shape. Both groups soften in 2020 and 2021, the big ones recover, the small ones keep sliding. Meanwhile the share of genuinely airy records, above the 2019 top quartile for energy over 8 kHz, is flat in every group, 23 to 27 percent every single year. The air was never the problem.

Why that is exactly what a bad room does

This is the part I should have predicted, because I have sat in those rooms.

A room with nothing on the walls does two things to you at once. Up top, early reflections off the desk, the window and the side walls arrive a few milliseconds after the direct sound and smear the 2 to 5 kHz region into something harsh and papery. You hear harshness that is not in the mix, and you cut it. Down low, the room's dimensions set up standing waves somewhere between about 100 and 400 Hz, and at the listening position some of those notes are twice as loud as they should be and some have vanished. You cannot make a decision about a region you cannot hear, so the buildup stays in.

Cut the presence, keep the mud. That is the pattern in the tables, in the right direction, in the right years, and it stuck in the group that never got the room fixed.

I noticed the half of this I expected and missed the half that matters. The bright room does not put brightness on the record. It takes the bite out, because you are fighting the room instead of the mix, and it leaves the weight in, because the room will not let you see it.

What I do about it in my own room

My room is treated, which is the boring answer and the only real one. Short of that, here is what has held up for me over twenty years of records and what I tell anyone mixing at home.

I do not high-pass my way out of low-mid trouble. A hard cliff at 100 Hz sounds like a plugin. I use a shelf, pull the region down a couple of dB and leave the floor there, just quieter, because the floor is where the record lives in a room. The low end is the rug. Fix the room, then put it back.

I check the balance in mono on a pair of small cubes before I trust anything. If the vocal and the kick still read when the stereo field is gone, the low mids are under control. If the mix turns to soup in mono, that is the 250 to 500 region telling on itself, and no amount of top-end polish fixes it.

I play the mix on a phone speaker. The phone has no low end and no air. All that is left is 300 Hz to 5 kHz, which is exactly the two regions in the tables above. If the vocal disappears on the phone, the presence got cut too hard. If the phone sounds boxy, the mud got left in. A six-dollar check that catches both mistakes an untreated room makes.

And I stopped mixing with the window behind the screen. That one is free.

Method

Every track was analyzed from audio by the same pipeline that powers the Sonic Analyzer; these numbers live in the table the analyzer benchmarks your scan against. Spectral energy is split into seven bands (sub below 60 Hz, bass 60-250, low mid 250-500, mid 500-2k, high mid 2-4k, presence 4-8k, air above 8k), measured on the track's highest-energy sections; brightness is the spectral centroid. Release dates come from the catalogue metadata; a track enters the year of its release, not the year we measured it. Size groups use the track's Spotify play count at the time of measurement, so "small" means a release that stayed small. Within-genre comparisons use each genre's own 2018-2024 median. All shares in the tables are medians or quartile counts, never means. The effects here are modest, a few percentage points on tens of thousands of records, and I have tried to describe them at that size.

Want to know where your own record sits against these tables? Run it through the analyzer. It is free, and it will tell you in about a minute whether you cut the bite or kept the mud.