Mastering Basics: Loudness, LUFS, and When You're Done

Mastering Is Quality Control, Not Magic

Mastering is the last step before a song leaves your hands: final EQ, final dynamics, final level, and a check that the file itself is technically clean. That's it. It is not a secret process that transforms a mediocre mix into a record, and it is not primarily about making things loud. A good master is mostly a good mix with the last five percent of polish and a delivery format that works everywhere.

That framing matters because the most common home mastering mistake is treating the limiter as the point of the exercise. People open a loudness meter, see a number that's lower than a commercial track, and push until the numbers match. The result is usually a master that's flatter, harsher, and — thanks to streaming normalization — not even louder in the listener's ears.

This guide covers the three things you actually need to understand: what LUFS measures, how loud your master should really be, and how to know when you're done. None of it requires expensive tools. Most DAWs (digital audio workstations — your recording software) now ship with a loudness meter, and free third-party loudness meters cover the rest.

Loudness Is Not Level: dBFS, RMS, and LUFS

Three different rulers get confused constantly, so let's separate them.

dBFS (decibels relative to full scale) measures peak level — how close the waveform gets to the digital ceiling. 0 dBFS is the maximum a digital file can hold. Peak level tells you whether you're clipping. It tells you almost nothing about how loud something sounds, because a single snare hit can peak at -1 dBFS while the track as a whole is quiet.

RMS (root mean square) is an average of the signal's energy over time. Closer to perceived loudness than peak, but it treats all frequencies equally, and your ears don't. We're much more sensitive to midrange than to deep bass, so two tracks with identical RMS can sound very different in loudness.

LUFS (Loudness Units relative to Full Scale) is the modern standard, defined by an international broadcast specification (ITU-R BS.1770). It's an energy average like RMS, but filtered to approximate human hearing — midrange counts more, sub-bass counts less — and it gates out silence so quiet gaps don't drag the number down. One LU (loudness unit) equals one decibel of level change. When a streaming service decides how much to turn your track up or down, LUFS is the measurement it uses.

The three LUFS readings on your meter

A loudness meter shows several numbers at once. They answer different questions:

The one that matters for "how loud is my master" is integrated LUFS, and it's only meaningful measured across the entire song. Looping your loudest chorus and reading integrated loudness gives you a misleadingly high number.

True peak: the ceiling that isn't where you think

Digital meters read sample values, but the actual analog waveform your converter reconstructs passes between samples — and it can swing higher than any individual sample. These are inter-sample peaks. A file whose samples never exceed 0 dBFS can still clip on playback, and lossy encoding (the compression Spotify, YouTube, and others apply to your file) makes it worse.

That's why true peak meters oversample the signal to estimate the real waveform, and why the standard advice is to set your limiter's ceiling to -1.0 dBTP (decibels true peak) for anything headed to streaming. Some platforms have suggested even more conservative ceilings for very loud masters. The half-decibel of "lost" loudness is inaudible; the encoding distortion you avoid is not.

How Loud Should Your Master Be?

Short answer: probably somewhere between -14 and -8 LUFS integrated, depending on genre — and the exact number matters much less than people think, because of normalization.

What streaming normalization actually does

Every major streaming platform measures your track's integrated loudness and adjusts its playback level so that everything on the platform sits at roughly the same perceived loudness. If you deliver a master at -8 LUFS and the platform targets -14, it simply turns your track down about 6 dB on playback.

Read that again, because it's the whole ballgame: extra loudness above the normalization target is thrown away. You paid for it with squashed dynamics, and the listener never hears the loudness — only the squashing. A -14 LUFS master and a -8 LUFS master of the same song, played back normalized, sit at the same volume. The more dynamic one usually sounds punchier, because its drum transients survived.

Commonly cited normalization targets, which are approximate and do change — treat the platforms' own documentation as the source of truth:

Two honest caveats. First, normalization isn't universal: some users disable it, some playback contexts skip it, and behavior differs between turning loud tracks down (nearly universal) and turning quiet tracks up (inconsistent — some platforms won't boost, or won't boost past the point of clipping). Second, the targets shift over time. That's precisely why chasing an exact number is a mistake.

So what number do you actually aim for?

Don't master to -14 LUFS because "that's what Spotify wants." Spotify doesn't want anything; it just turns the knob. Master to what the song needs, then check that you're in a sane range:

• Dense modern pop, rock, hip-hop, EDM: most commercial releases land somewhere around -10 to -8 LUFS integrated, some hotter. This is partly competitive habit left over from the loudness war — the decades-long escalation of master levels — and partly aesthetic: some genres genuinely want that pressed, saturated density.
• Singer-songwriter, acoustic, jazz, classical: -16 to -12 LUFS is common, and dynamics are the point.
• If you're unsure: landing between -14 and -10 LUFS with a -1 dBTP ceiling is a safe, professional-sounding zone for almost any home release.

The honest state of expert opinion: engineers disagree about whether there's any reason left to master hot. Some argue normalization has made loudness maximization pointless; others point out that non-normalized playback still exists and that heavy limiting is now baked into the sound of certain genres. Both positions are defensible. What nobody credible argues is that loudness alone makes a master better — so when in doubt, err dynamic.

The most reliable calibration isn't a number at all. Loudness-match a commercial reference track in your genre against your master — turn the reference down until both read the same short-term LUFS — and compare. If yours sounds thin or wildly less dense at matched loudness, the problem is the mix or the tonal balance, not the limiter setting.

Mastering at Home: A Minimal Chain That Works

You do not need a mastering suite. You need a clean mix, a meter, and restraint.

Before you touch the master bus

1. Fix it in the mix. If the low end is boomy or the vocal is harsh, go back to the mix session. Mastering EQ moves are broad strokes; surgical problems belong upstream. And if your mixes routinely sound different everywhere you play them, sort out why your mixes aren't translating before mastering anything — a master of an untranslatable mix is just a louder untranslatable mix.
2. Leave headroom. Export your mix with peaks somewhere around -6 to -3 dBFS and no limiter on the mix bus (bus compression that's part of your sound can stay). There's no fidelity magic in those exact numbers; the point is simply that nothing clips and the mastering limiter isn't fighting a squashed input. Don't "normalize" the exported file either — that just eats the headroom you deliberately left.
3. Export at full resolution. Master from a 24-bit (or 32-bit float) file at your session's native sample rate — the reasoning is covered in our guide to sample rate and bit depth. Don't master from an MP3.

The chain itself

A sensible home mastering chain, in order:

1. Corrective EQ (broad, small). Cuts or boosts of 0.5–2 dB over wide bandwidths. If you're reaching for more than about 3 dB, that's mix feedback, not a mastering move. A high-pass filter below the lowest musical content (often 20–30 Hz) cleans up inaudible rumble that eats limiter headroom.
2. Optional glue compression. A bus compressor with a low ratio (around 1.5:1 to 2:1), slow attack, and 1–2 dB of gain reduction on the loudest sections. It should be nearly invisible. Skip it entirely if you don't hear an improvement — plenty of good masters use no compressor at all.
3. Limiter last. Set the output ceiling to -1.0 dBTP. Raise the input gain (or threshold) until the loud sections show roughly 2–4 dB of gain reduction on peaks. Then listen — to the drums especially. When the kick and snare start to feel blunted, or the track "breathes" or pumps, back off 1–2 dB. Where the limiter starts hurting is your loudness ceiling for this song, whatever the LUFS meter says.
4. Meter and verify. Play the whole track through with a loudness meter running. Note the integrated LUFS and true peak. If the integrated number lands somewhere reasonable for your genre and nothing sounds strained, you're in good shape.

Loudness maximizers, multiband dynamics, stereo wideners, and "mastering-grade" exciter plug-ins are all things you can add later, if ever. Each one is easier to hurt a master with than to help it. The chain above, used gently, is how a very large share of respectable independent releases actually get finished.

The one-fader sanity check

Limiters make things louder, and louder always sounds better on first listen — that's a documented perceptual bias, not an opinion. Defend against it: after setting your limiter, pull its output down until the processed track matches the unprocessed mix in loudness, and A/B them. If the "mastered" version doesn't sound at least as good at matched volume, your processing is subtracting, not adding.

Monitor at Sane Levels — Your Ears Are the Instrument

Mastering decisions are made by ear, and ears compress and dull with fatigue. Two practical rules:

• Work at moderate volume. Conversational-to-slightly-loud — roughly the level where you could still talk over the music — is where your hearing's frequency response is most balanced and where you can work for hours. Check loud briefly; don't live there.
• Respect exposure limits. Sustained listening at high levels causes permanent hearing damage; occupational guidelines treat around 85 dB SPL as the level where exposure time starts needing limits, and home studio sessions can easily exceed that without feeling extreme. Take a 5–10 minute silence break every hour. Everything sounds clearer after one, which tells you something.

A fatigued decision at 100 dB is how masters end up dull and over-limited. Quiet, rested ears are a competitive advantage that costs nothing.

When You're Done: A Finish-Line Checklist

"Done" is a decision, not a feeling, so make it mechanical:

• Full-length listen, no touching anything. Start to finish, at moderate volume. Note problems; don't fix them mid-playback.
• Loudness-matched reference check. Your master against two or three commercial tracks in the genre, levels matched by ear or meter. You're comparing tonal balance and density, not volume.
• Small-speaker and earbud pass. Phone speaker, cheap earbuds, a car if you have one. You're listening for a vocal that disappears, bass that vanishes entirely, or harshness that only shows up on bright playback.
• Meter sanity check. Integrated LUFS in a sensible range for the genre; true peak at or below -1.0 dBTP; no clipping anywhere.
• Top and tail. A short fade or clean edit at the start and end, no dead air, no cut-off reverb tail.
• Sleep on it once. One overnight gap catches more mistakes than any plug-in. If it still holds up tomorrow, ship it.

What "done" does not require: matching a commercial track's loudness number, using any particular plug-in, or the mix feeling perfect. Records get finished by people willing to stop.

The takeaway: master for a sane integrated loudness (-14 to -10 LUFS is safe for most genres), keep true peaks under -1 dBTP, stop pushing the limiter the moment the drums complain — and let the streaming services worry about the volume knob, because they will anyway.