Two in the morning is the only honest time to cut a master. The room is quiet, the phone has stopped, and you can finally hear what the mix actually is instead of what you hoped it would be.
I spent years doing this on WIR Records. Not as a mastering engineer with a plate on the door and a Buchla in the corner, but as the label hand who ended up with the last stage because nobody else wanted to sit there at 2 AM arguing with a LinnDrum that keeps poking through a vocal. What follows is the unglamorous stuff. The processing order I stopped fighting, the software that broke on me, and the things about mastering I wish someone had said flat out instead of dressing up as a framework you can steal.
Mastering with intention, not with a preset list
You can memorize a hundred plugin chains and still ruin a track. A mastering chain is a set of decisions, not a shopping list. According to the framework laid out at mastering.com, the whole job comes down to three goals: make the track louder, make it sound better, make it translate across systems. Everything on the bus should serve one of those three. If it doesn’t, cut the plugin loose.
That framing saved me more wasted nights than any specific tool. At 2 AM your ears are tired and the temptation is to add, add, add. The three rules push back. Why is this EQ move here? Which of the three goals does this limiter earn? Most of my worst masters were the ones where I broke those rules and processed because I could.
The chain I actually landed on
Order matters less than purpose, but I keep coming back to a sequence that mirrors what mastering.com calls intent-driven processing. Cleaning EQ first, then glueing compression, then the limiter stage, with treatment in between that I’ll get to.
Cleaning EQ before anything else
The first EQ is not for tone. It’s for removing what shouldn’t be there. Broad Qs, small moves, transparent plugin. If the high mids bite, pull them back a little. If the sub is heavy in one spot, shave it. This is maintenance, not art. Get it wrong and everything downstream magnifies the mess.
I keep a de-esser in reserve for this stage too. When the Ss poke or the cymbals get glassy, a dedicated de-esser reacts to peaks that a static EQ can’t. But keep it subtle. You’re working on the whole file, and one over-eager de-ess can dull a top end that took the artist three days to get right.
Glue before color
A glueing compressor is what separates a mastered track from a pile of stems that happen to be in key. The telltale sign of an unmastered mix is a lack of glue, the mastering.com guide nails that. I run a slow, gentle setting, maybe 1 to 2 dB of movement, just enough to make the low end and the vocal breathe together. Makeup gain handled by the plugin itself gives me a fuller result than pushing the output manually.
You do not need two compressors. The guide is honest about this: there are no hard rules, and two only makes sense if you’re splitting clean glue from a colored one. I usually split them anyway because I like a clean glue stage and a separate color stage, but I’ve cut plenty of records with one good compressor doing both jobs.
Clipping, stacking limiters and the loudness lie
Here’s the thing nobody told me until I’d already learned it the hard way. Slamming a single limiter is not how you get loud without damage.
A studio-floor tip from a Studio One forum thread that stuck with me: shave 2 to 4 dB of transients with a clipper before the limiter, then let the limiter push the level. The poster’s point was that taking the clipper out and compensating with the limiter gets you pumping or distortion. I’ve heard that exact difference on my own bus more times than I care to count. The clipper takes the peaks the limiter would otherwise fight, so the limiter works less and sounds cleaner.
The forum also suggests stacking two limiters to reduce the load on each. I resisted that for years because it felt like a hack. It isn’t. Splitting the same amount of gain reduction across two stages keeps each one out of its worst range, and you get the timbre of both units. The sageaudio chain agrees, using a FabFilter Pro L2 on its modern setting for the clean final catch. That L2 is transparent enough that I stopped arguing with the second limiter and just left it in.
The loudness war is over in the sense that the streaming platforms are fighting back. The mastering.com guide says it plainly: you can no longer just throw a limiter on and shove the output as hot as it will take. For WIR, the target was never maximum LUFS. It was the point where the track stopped getting better and started getting louder. That point arrives earlier than your ego wants.
The color stage I kept coming back to
Saturation at 2 AM is a mood, not a spec. I used a tube stage for fill, particularly when a mix had a gap in the mid range that EQ couldn’t fix without carving. The Saturn 2 tube setting is what I reached for on the masters that needed body, and the sageaudio chain confirms the choice: the tube algorithm fills in the spectrum with harmonic distortion and compression that feels natural on a full mix. Splitting it into three bands gives control over which range takes the color, though the modulation and filters stay on the middle band and don’t follow to the new low and high bands. That’s a limitation, not a bug, and it suits mid-range work anyway.
The linear phase option is worth the extra CPU if you’re on a deadline that has minutes to spare. It keeps the distortion out of the artifacts. And if a master’s highs were harsh but I still wanted them present, a warm tape saturator at 30 ips was my fallback. The faster speed keeps the sound clean and pulls the lows down a touch, which is exactly the trade I wanted on thin pop masters.
Stereo, phase and the wide-but-wrong trap
Widening the master is one of the most tempting moves and one of the easiest to get wrong. The mastering.com chain warns that most stereo expanders use the Haas effect, and pushing them too far creates phase issues. I’ve heard records that sound enormous on headphones and collapse to nothing on a mono club system. That collapse is phase cancellation, and it comes from a wide move that felt good at 2 AM.
My rule: check mono before you commit. If the track loses anything important, back the width off. A great mix will already be wide. Your job is to protect it, not to re-imagine it.
The software layer nobody writes about
WIR ran on borrowed time and old drivers. Half the 2 AM sessions weren’t about the chain at all. They were about a plugin that loaded but didn’t pass audio until the ASIO buffer was reconciled, or an interface that needed a driver version pinned to a specific firmware before the meters behaved. If your audio interface drivers are out of step, no amount of Pro L2 top-end will save the session. The software layer underneath is where masters die.
My suggestion, born of too many failed print attempts: before you build a chain, verify the boring stuff. Confirm the driver version matches the interface firmware. Set the buffer where you actually want it and check the latency reading, not just the slider. Then leave the machine alone. Half the instability I chased was caused by tweaking a working system at 1 AM because I couldn’t hear the difference I wanted and blamed the platform. It was never the platform.
What I’d tell the next person in the chair
Know why you’re processing before you process. Use the three goals. Clean before you color, glue before you compress for color, clip before you limit, and stack limiters rather than asking one to do everything. Check mono. Leave the drivers alone once they work.
And print the mix that sounds right, not the one that hits the loudness target you set at midnight. The 2 AM ear lies about loudness. Trust the version you’d play for a friend the next morning, when the room is bright and the coffee is fresh. That’s the master.