On Tuesday I showed what mastering moved: loudness, true peak, clipping. Today is the other half of the same data, and to me it's the more interesting half. Across 78 masters and 69 unmastered files, the things that make a mix sound like that mix barely changed.
Same setup as Tuesday: two groups of songs, not before and after pairs of one song. So these are typical numbers for each group, not a controlled test. No song titles or customer names.
The numbers that didn't move
These are medians, the middle song in each group.
Stereo correlation (how similar left and right are, where 1 is mono): 0.82 before, 0.82 after.
Mono fold-down change (how much quieter the song gets summed to mono): -0.53 dB before, -0.53 dB after.
Stereo width (how wide the mix measures): -10.0 dB before, -10.1 dB after.
Tonal balance, share of energy by band: sub 17% to 16%, low 42% to 42%, mid 34% to 35%, high 3.8% to 3.8%.
Loudness range (how much loudness moves across the song): 5.2 LU before, 5.0 LU after.
Every tonal balance share landed within about a point of where it started. Stereo correlation and mono fold-down didn't move at all.
What that means for your song
Mastering didn't widen anyone's stereo field. It didn't reshape anyone's tone. It didn't flatten or open up the song's dynamics. It made the song consistent and streaming-ready, and it left the artist's choices where the artist put them.
That's how I want it. A master that changes the width and the tone is a remix, and nobody asked me for a remix. If your song is too narrow, too dark or too bright, the fix sits earlier: in the generation, the stems or the mix. Mastering finishes what's there.
One honest note on dynamics. Peaks did come down relative to loudness: the median peak to loudness ratio went from 11.5 dB to 10.7 dB. That's what limiting does. But loudness range, how much the song moves between its quiet and loud sections, stayed put at 5.2 vs 5.0 LU. One master has a loudness range of 12.6 LU. That's a dynamic song I left dynamic, not a fault.
Mono is still safe
Plenty of people still hear songs in mono: a phone speaker, a Bluetooth speaker, a bar system. Every file in both groups, mastered or not, cleared my mono checks, and the median song lost about half a dB when summed to mono. Mastering didn't make that worse.
The one thing mastering can't add back
28% of the masters and 22% of the unmastered files roll off sharply somewhere between about 17.7 and 18.6 kHz. That ceiling comes from the AI source. Mastering can't put back detail that was never in the file.
The master share is higher only because these are different songs in each group. It isn't something mastering did.
That range is hard for a lot of listeners to hear at all, so I wouldn't panic about it. But it's worth knowing that the top end is set at the generation and export step. Export the best file your tool gives you, a WAV rather than an MP3, before anything else happens to it.
Check your own file
Track Lab is a free checker I built. Drop in a file and it measures loudness, true peak, clipping and more. It's measurement only. It can show you a number, but deciding what to do about it still takes ears.
If you want a master that makes your song consistent and streaming-ready without turning it into somebody else's mix, that's the work I do, myself, start to finish.
