I Master My Own Catalog on Purpose
Nobody hears my records more than I do. By the time a track is done I have listened to it hundreds of times — in the car, on the phone speaker, on studio monitors at 2 a.m. That much exposure is a curse for mastering, because your ears stop hearing the song and start hearing the mistakes you already know are there. You start chasing ghosts, pushing the loudness up to feel something, and you regress the master a little every pass without noticing.
The doctrine below is the discipline I built to protect against my own ears. It is what kept a ten-track album like DARK I: Outwitting the Devil sounding like one record instead of ten different mastering moods. It is not gear worship and it is not a loudness-war smash. It is a reference-matched, never-regress process that lands every track at −14 LUFS and stops there.
The Reference Is the Whole Game
I never master to a number in a vacuum. A number tells you how loud something is, not whether it sounds right. So before I touch a single band, I pick a finished, commercially released track that already sounds the way I want mine to sound — my internal reference is codenamed ICEMAN — and I master toward it.
Reference-matching in plain English: you measure the tonal balance of the reference — how much low end, how much midrange, how much air — and you nudge your own track to sit in the same neighborhood. You are not copying the reference's arrangement or its vibe. You are borrowing its balance as a target, because a well-mastered commercial record is a calibration tool sitting right there for free.
The moment you have a fixed target, the guesswork collapses. Instead of "does this sound good?" — a question your tired ears cannot answer honestly — you ask "is this closer to ICEMAN than the last pass was?" That is a question a measurement can answer.
Separate Before You Touch Anything
Here is the mistake most bedroom masters make: they slap a chain across the whole stereo mix and squash the vocal, the sub, and the beat with the same medicine. A vocal needs de-essing and repair. A sub-bass needs to be left almost entirely alone. Treating them as one signal means every fix you make to one is damage to the other.
So I separate first. I run BS-RoFormer stem separation locally on the Mac — Apple Silicon, MPS acceleration — to pull the vocal off the instrumental. Then repair happens only on the vocal stem: click removal, breath control, sibilance, the surgical stuff. Critically, I reinject only the exact repair complement — the difference the repair made — back into the track. I never re-color the sub, never re-EQ the beat, never let a vocal fix leak into the low end. The instrumental keeps the weight the mix engineer gave it. The vocal gets clean. Nothing bleeds.
The 31-Band Match, Then −14 LUFS
With a clean signal and a target, the match itself is a 31-band spectral comparison against the reference. Thirty-one bands is enough resolution to correct real tonal drift — a boxy 300 Hz, a dull top end, a bass region that is 2 dB hot — without carving the track into something artificial. I move toward the reference's curve, not onto it. The goal is my record with the reference's discipline, not a clone.
Only after the tone is right do I set loudness. I normalize to −14 LUFS integrated — the level the streaming platforms are actually built around. Master hotter than that and Spotify, Apple Music, and YouTube turn you back down anyway, except now they turn down a squashed, lifeless master instead of a punchy one. −14 LUFS is not being quiet. It is being honest about where the music will actually be heard, and keeping the dynamics that make a record hit.
Never Regress
This is the rule that names the whole doctrine, and it is the most important sentence I will write here: every pass must score at least as well against the reference as the pass before it, or it does not ship.
The engine that enforces this is a script I call reference_master.py. It scores each candidate master against the reference and against the previous version. If a new pass is measurably worse — louder but duller, closer in one band but further in three — it gets rejected automatically. I cannot talk myself into a regression at 2 a.m. because the score will not let me. A master only leaves the room when it is what I call DAJAI APPROVED — better than every version that came before it, verified, not vibed.
That single guardrail is the difference between a catalog that gets better over time and one that drifts. Human ears fatigue. A never-regress rule does not.
Why an Independent Artist Should Own This
You do not have to master your own music. But you should understand that owning the master chain end to end is leverage no distributor and no third-party engineer hands you. DARK I is ten tracks, out April 7, 2026 on Apple Music and Spotify, and every one of them went through this exact process under my own hands. When I want to re-master, re-sequence, or re-issue, I do not file a ticket and wait. I open the session.
Reference-matching, stem separation, a 31-band target, −14 LUFS, and a never-regress rule — none of it requires a boutique studio. It requires a reference you trust, a way to measure against it, and the discipline to let the measurement win over your ears. That is a workflow, not a budget. And for an independent artist, workflow is the whole edge.
FAQ
What LUFS should I master to for streaming?
−14 LUFS integrated is the practical target. Apple Music, Spotify, and YouTube all normalize playback, so mastering hotter than that just gets turned back down — except you have already sacrificed dynamics to get there. Master to −14 LUFS and keep the punch.
Do I need expensive plugins to reference-master?
No. Reference mastering is a method, not a plugin. You need a commercial track you trust as a target, a way to measure your tonal balance against it, and the discipline to move toward it. The gear that does this ranges from free to boutique — the target and the discipline matter far more than the price of the tools.
What does "never regress" actually mean?
It means every new pass of a master must measure at least as well against your reference as the previous pass, or you discard it. It protects you from the classic mastering trap: pushing loudness or chasing one band while quietly making the overall master worse. A score, not your tired ears, decides whether a version ships.
Can I master my own music as an independent artist?
Yes, and there is real leverage in it — you can re-master and re-issue without waiting on anyone. The catch is that your ears fatigue on your own material, so you need an external target (a reference track) and an objective check (a never-regress rule) to keep yourself honest. With those two guardrails, owning your master chain becomes an edge instead of a liability.