The Complete Guide to Audio Mastering

TL;DR: Audio mastering is the final stage that decides how a song translates everywhere it gets played. Start with a clean import (peaks at -6 to -3 dBFS, integrated loudness around -20 to -16 LUFS), use one to three level-matched references, then build a small, purposeful chain: surgical EQ, tonal EQ, light compression if needed, gentle saturation, careful M/S width, de-essing, and a transparent limiter at -1.0 dBTP or lower. Most great masters are two or three precise moves, not a stack of plugins.

What Is Audio Mastering, Really?

If mixing is where your song learns to walk, mastering is where it learns to travel. It’s the final 5% of the process that decides how the other 95% will be heard, on phones, in cars, on club systems, through cheap earbuds, and in the playlists your track will live beside.

My job in audio mastering isn’t to “fix” what’s broken. It’s to reveal what’s already great, protect your dynamics, and make the music translate everywhere without losing its character. In this guide, I’ll walk you through how I approach a song, from the first listen to the final export. I’ll share the judgment calls I make, the numbers I care about, and when to reach for different tools. This is written for beginner and intermediate producers, the same way I’d guide you in a session.

What Mastering Actually Is (And Isn’t)

Mastering is the last creative and technical stage before release. I align the tonal balance, shape the dynamics, control peaks, and prepare files that meet delivery specs for streaming or physical formats. It’s not a second mix. It’s not a place to throw a plugin chain at every song. Most great masters are a few precise moves in the right order. For a deeper introduction to the concept, see my guide on what audio mastering actually is.

When you bring me a mix, I’m listening for:

• Clarity in the midrange (where music “lives”)
• Low-end shape and tightness
• Top-end smoothness (air vs. harshness)
• Stereo width and depth
• Transient definition
• How the track feels next to realistic references in the same lane

How I Start a Session

I begin with a clean import of your final stereo mix at its original resolution: 24-bit or 32-bit float, 44.1 to 96 kHz. I want healthy headroom.

What I’m looking for:

• Peaks around -6 to -3 dBFS
• Integrated loudness roughly -20 to -16 LUFS
• No clip lights anywhere

If your mix bus has a limiter or clipper that’s purely for loudness, I’ll usually ask you to remove it. If it’s part of the sound (creative clipping on the snare, for example), we keep it and adjust the mastering approach. Getting the mix prepared properly is half the battle, which is why I’ve written a separate guide on how to prepare your mix for mastering.

Loading Reference Tracks

I load one to three reference tracks that match your aesthetic. Level-match them to my working loudness within about half a dB. This is crucial. Louder always feels better, so we remove that bias. When I A/B, I’m comparing sections:

• The energy and shape of the drop or chorus
• The vocal brightness relative to cymbals and synths
• The low-end contour in the kick and bass relationship
• The feeling of depth vs. width

The other thing I do early: make sure my monitoring path is level-matched for instant, fair comparisons. A simple post-chain trim lets me flick the whole chain on or off at the same perceived loudness. It keeps me honest.

Gain Staging and Why It Matters

Good audio mastering is mostly good gain staging. If every processor sees a healthy, predictable signal, it behaves musically and meters tell the truth.

Here’s what works:

• Pre-chain peaks around -6 to -3 dBFS
• Mix loudness around -20 to -16 LUFS integrated
• Limiter ceiling at or below -1.0 dBTP to reduce inter-sample clipping on consumer DACs

Those numbers aren’t religion. They’re guardrails. If you stay near them, your tools are less likely to lie to you, and the last stage (the limiter) has room to do its job without chewing up transients. For a deeper breakdown of why this matters, see my guide on what gain staging is and how to do it right.

Building the Chain (And Why Order Matters)

I don’t start with a “template.” I sketch a chain for the song in front of me. Many times, it’s as little as two or three processors. Other times it’s more, but each block has a job. My typical order, adjusted for the specific mix, looks like this.

1) Surgical EQ

First I fix what hurts. Narrow cuts for ringing resonances, a touch of low-end cleanup if there’s mud or sub-rumble that stomps on headroom. I’m careful with high-pass filters: they can solve problems, but they also shift the feel of the track. If I don’t need one, I don’t use it.

2) Tonal EQ

Then I paint with broad strokes: 0.5 to 1.5 dB shelves or wide bells to bring the record closer to the emotional target. If the reference has a bit more “smile” at 12 kHz, I’ll try a gentle top shelf. If the vocal sits slightly forward on the reference, a wide 1-2 kHz lift might do it. These are small moves. Many masters come alive with two well-placed half-decibel adjustments.

3) Compression (Only If the Song Asks for It)

If the mix needs glue, I’ll try gentle bus compression: 1-2 dB of gain reduction, slow attack to keep transients, medium release so the groove breathes. If the mix already has great internal movement, I skip compression. It’s amazing how often “nothing” is the right setting here.

4) Harmonics and Density

This is where subtle saturation can add body, cohesion, or a touch of “record” to a clean digital mix. In-the-box, I’ll reach for tasteful transformer or tape-style tools and watch the upper mids like a hawk. Used sparingly, these can add the glue a mix needs without making it sound processed.

5) Stereo Field (And M/S Judgment)

Width is addictive, and dangerous. If I want a touch more size, I may use mid/side EQ to gently lift high-end air in the sides or tuck a harsh midrange in the side channel. The test is always mono: if the chorus collapses or the kick/bass phase gets weird, it wasn’t width, it was a trick.

6) De-harshing and De-essing

Sometimes the harshness isn’t volume. It’s spectral shape. A broadband de-esser or a narrow dynamic dip around 2-5 kHz can take the edge off, especially on bright vocals and cymbals. The goal is polish, not dullness.

7) Limiting (Where Loudness Meets Taste)

I set the ceiling at or below -1.0 dBTP and bring up level until I’m trading the least punch for the loudness I need. I watch short-term LUFS in the loudest section to judge where the chorus sits. If transient snap starts to blur, I back off and let the record breathe. For a list of the limiters I actually trust, the cheat sheet above includes my Top 3 Transparent Limiters.

Loudness (And How I Aim)

Most services normalize playback toward the mid-teens LUFS integrated, often around -14 LUFS, but implementations vary and users can disable normalization. Two truths can coexist: you should respect the reality of normalization, and you should also make your master sound great at its native level.

My default guardrails are simple:

• Limiter ceiling at or below -1.0 dBTP
• Integrated loudness that lands comfortably in the neighborhood services expect

Personally, when the style calls for it, I will push to about -10 LUFS at the loudest section of the song while doing everything I can to keep transient integrity and a healthy crest factor. That might mean accepting a slightly lower integrated number if it sounds more alive, or taking the time to shave 0.3 dB in a harsh upper-mid area so the limiter works less hard.

Streaming Platform Targets

Here’s what you need to know about the major platforms:

In-the-Box Tools That Work

You don’t need hardware to master well. A solid in-the-box toolkit can get you there:

• One clean, fully parametric EQ for surgery and small tonal moves
• One musical or modeled EQ for broad color when needed
• A transparent bus compressor that can do 1-2 dB without pumping
• A tasteful saturator or tape/transformer for density (used sparingly)
• A modern limiter with true-peak control and clear metering
• Metering that shows LUFS (integrated and short-term), true peak, a spectrum analyzer, and ideally a vectorscope

The trick isn’t the brand. It’s the discipline. Keep your monitoring honest, level-match every comparison, and do less than you think you should. Most “budget problems” are solved with taste, not plugins. For specific recommendations, see my guide on the best mastering plugins for home studios.

Reference Tracks Without Lying to Yourself

References are a compass, not a destination. Level-match them within about 0.5 dB of your working loudness, then compare fast: chorus to chorus, intro to intro.

Listen for:

• Low-end shape: Is your kick/bass relationship tight or foggy?
• Vocal placement: Forward, tucked, or riding the band? Any harshness around 2-5 kHz?
• Top-end air: Expensive “sheen” versus brittle “spray”
• Width vs. depth: Do you feel a front-to-back stage, or just left-right?

If your master feels smaller, resist the urge to “turn up the sides.” Often the answer is a 0.5 dB tonal correction, not a width trick.

Common Problems I See (And How I Fix Them)

Harsh Top End

Usually a combination of hot 2-5 kHz and splashy 10-12 kHz. I’ll try a gentle M/S move (soften the side channel in that presence zone) or use a high-frequency compressor to shave the spikiest transients without dulling the tone. Sometimes the fix is earlier in the chain: reduce a narrow upper-mid resonance before boosting air.

Flat Low-End

If the sub is washing over everything below 80 Hz, I’ll do a small, narrow cut where the room is blooming, or use a targeted, slow-release multiband to hold the very bottom steady while the punch lives around 100 Hz. The test is always kick definition: does the chest hit come back without thinning the song?

Smeared Transients

This is often limiter over-work or a compressor attack that’s too fast. I’ll ease the limiter, raise its ceiling slightly, or split a dB of loudness across two gentler stages. Sometimes simply reducing 0.5 dB of upper-mid harshness lets the limiter work cleaner and the transients wake up again.

Over-Processing

If every plugin looks “busy,” I bypass the entire chain and rebuild. Most great masters are the result of two or three surgical decisions, not a dozen clever tricks. For more pitfalls to avoid, see my guide on 10 audio mastering mistakes most producers make.

When Stem Mastering Makes Sense

If the mix is close but one element keeps fighting (say the vocal is 1 dB too hot, or the kick masks the bass), stem mastering can be a surgical middle ground without reopening the entire mix session. If you go this route, keep stems phase-aligned, with the exact same start time and format as the main mix. Then treat it like mastering, not a remix: minimal, goal-driven moves that protect the original intent.

Deliverables That Won’t Come Back to Haunt You

For streaming and digital distribution, I export a 24-bit WAV at the project’s sample rate (44.1 or 48 kHz are most common; 96 kHz when the project calls for it). If a 16-bit version is needed, I apply shaped dither as the final step and export a separate 16-bit WAV. For CDs, I’ll generate a DDP with CD-Text and ISRC where applicable.

File naming: I name files clearly (Artist_Song_Master_v01.wav) and increment with every revision so no one is guessing which file is “the one.” I also document the final readings: integrated and short-term LUFS, true peak, and any notable chain moves. It makes revisions faster and future releases more consistent.

Targets Worth Remembering

Here are the numbers that matter:

• Pre-chain peaks: around -6 to -3 dBFS
• Mix loudness: roughly -20 to -16 LUFS integrated
• Limiter ceiling: ≤ -1.0 dBTP
• Streaming reality: many services normalize around -14 LUFS integrated
• My preference when the genre wants it: about -10 LUFS at the loudest section, with transients intact

Key Takeaways

• Audio mastering is the last 5% that decides how the rest of the work translates everywhere.
• Start with healthy headroom: peaks at -6 to -3 dBFS, integrated loudness around -20 to -16 LUFS.
• Use level-matched references (within 0.5 dB) and compare section by section.
• Build the chain for the song, not from a template. Two or three precise moves beat ten clever ones.
• Set your limiter ceiling at -1.0 dBTP or lower to handle inter-sample peaks.
• Streaming platforms normalize to roughly -14 LUFS, so prioritize transient integrity over raw loudness.
• Keep low frequencies centered and always check mono compatibility.
• The ear is the final meter. Numbers are guardrails, not goals.
• Deliver 24-bit WAVs at the project’s sample rate, dithered down only if needed.

Related Articles

If you found this helpful, check out these guides:

• What Is Audio Mastering – The deeper introduction to what mastering is and what it isn’t.
• Audio Mastering for Beginners – A beginner-friendly walkthrough of the full process.
• How to Prepare Your Mix for Mastering – Get your mix ready before you master.
• 10 Mastering Mistakes Most Producers Make – Common errors that hurt your masters.
• Best Mastering Plugins for Home Studios – The tools I’d actually recommend buying.

Ready to Go Deeper?

This guide covers the principles. If you want to see them applied across full sessions, with my actual chain decisions, downloadable files, and the judgment calls I make on real projects, check out the Pro Audio Mastering course. It’s the same process I’ve used on releases for Armada Music and countless client projects.

Learn more about the Pro Audio Mastering Course →

About the Author

I’m Adrian Milea, a Mixing & Mastering Engineer, Music Producer, and Educator with 25+ years of experience. My work has been released by Armada Music and featured in DJ Mag. I run adrianmilea.com to help home producers get professional results without the guesswork.

Last updated: May 2026

Until next time, Stay Cool and Make Great Music!

-Adrian Milea