Mastering is the final stage of audio production where you optimise a mix for loudness, tonal balance, stereo width, and distribution across streaming platforms. A solid mastering chain runs EQ, compression, stereo widening, and a limiter in sequence, with subtle adjustments of +/- 1-2 dB and a final target loudness between -14 and -9 LUFS depending on genre and platform.

Good mastering is invisible. Nobody hears it and thinks “great mastering.” They just feel like the song belongs on the radio.

Bad mastering, though? You hear that instantly. The low-mids turn to soup. The high-end bites. The whole track sounds like it’s fighting itself at 2am through earbuds.

We’ve sat through both sides of that enough times to write this guide with some opinions. This isn’t a basic “slap a limiter on it” walkthrough. We’re covering the full signal chain, the exact numbers that matter, the metering tools professionals use, genre-specific targets, AI workflow integration, and the specific problems that other guides skip because they’re writing for beginners who just want something loud.

You want something loud AND right. That’s what we’re here for.

What Does a Mastering Signal Chain Actually Look Like?

The order matters as much as the tools. A standard mastering chain runs like this: high-pass filter, broad EQ correction, compression, stereo width processing, limiting. Sometimes a second, surgical EQ pass sits after compression. Harmonic saturation can live anywhere between compression and limiting depending on what you’re doing to the transients.

Don’t treat this as fixed gospel. It’s a starting framework.

High-Pass Filter First

Always. Cut everything below 20-30 Hz. There’s nothing useful down there in most genres, but a wandering subsonic rumble will eat your headroom and chew through your limiter like it has something to prove.

In iZotope Ozone 11, we run a Brickwall high-pass at 20 Hz on every session before touching anything else. It’s not exciting. It saves your master every time.

EQ: Broad Strokes, Small Moves

Professional mastering EQ operates in a range of +/- 1-2 dB maximum for corrective work. If you’re pulling 6 dB at 300 Hz, the mix needed to come back for revision, not mastering.

Broad Q settings, typically 0.5 to 1.0, work for tonal shaping. Narrow Q settings, 2.0 to 4.0, work for problem frequencies you couldn’t avoid inheriting. The Fabfilter Pro-Q 3 in Linear Phase mode is what we reach for when we need surgical precision without phase artifacts. Mid-side processing on a high shelf, boosting maybe 0.8 dB at 12 kHz on the sides only, adds air without touching the mono centre image.

Compression: Control, Not Crush

Mastering compression is about glue, not gain reduction. A 2:1 ratio with a slow attack around 30-50ms and a release around 150-250ms lets transients breathe while controlling the overall dynamic envelope. Gain reduction beyond 3 dB in mastering compression is a red flag. If your limiter is doing 6+ dB of reduction on top of that, the mix needs work, not more limiting.

For low-end control specifically, a multi-band approach at 2:1 on the sub-bass region (below 100 Hz) keeps kick drums from triggering the whole master into pumping. We love the for dialling this in without creating problems in the crossover regions.

What Loudness Numbers Should You Actually Hit?

Here’s where guides get vague and we’re going to be specific.

2024 Billboard Hot 100 data shows the majority of charting tracks mastered between -9 and -7 LUFS integrated. Those are loud records. Pop, hip-hop, EDM records targeting mainstream streaming sit there.

Platform targets are different:

Spotify normalises to -14 LUFS. Apple Music normalises to -16 LUFS. YouTube normalises to -14 LUFS. Tidal normalises to -14 LUFS.

What this means in practice: if you deliver a track at -8 LUFS integrated, Spotify turns it down 6 LU on playback. You didn’t gain anything from the extra limiting. You just lost dynamic range for no reason.

Our recommendation is genre-dependent. Deliver at -14 LUFS for acoustic, jazz, classical, and podcast-adjacent content. Deliver at -10 to -9 LUFS for commercial pop, hip-hop, and EDM. Master at -7 LUFS only if your A&R explicitly asks for it or you’re cutting vinyl.

True peak should never exceed -1.0 dBTP for streaming. Most platforms will introduce inter-sample peak distortion on lossy encoding if you push above that. Set -1.0 dBTP as the ceiling in your limiter and don’t touch it.

Metering Tools You Need Open

One meter isn’t enough. Professional mastering sessions run at minimum three visual feeds simultaneously.

An integrated LUFS meter tells you where the whole track lands. A momentary loudness meter (using a 400ms window) shows you short-term peak activity. A true peak meter watches for the encoder headroom limit. Youlean Loudness Meter 2 (free) covers all three in one window and we’d recommend it for anyone not yet running a full suite.

For short-term LUFS, the reading uses a 3-second sliding window. This is what catches a chorus that spikes 4 LU above the verse, which is the reading that tells you whether your compression is working or just decorating the signal chain.

How Do You Prepare a Mix File for Mastering?

This is the most skipped section in every guide and the most frustrating to fix after the fact.

Deliver 24-bit WAV files at your project sample rate. Do not bounce at 16-bit. Do not convert sample rate on bounce. If your project is 48kHz, deliver 48kHz. If it’s 44.1kHz, deliver 44.1kHz. We have had clients send 128kbps MP3 files for “mastering.” We sent them back. Don’t be that person.

Headroom peaks should sit between -3 and -6 dBFS on the mix bus output. This gives the mastering chain room to work without hitting the ceiling before the limiter even loads.

If you’re working at 44.1kHz, applying oversampling inside your limiter (most modern limiters including FabFilter Pro-L 2 and Weiss DS1-MK3 plugin emulations support 2x to 8x oversampling) effectively runs true peak detection at a higher resolution. At 96kHz native, oversampling matters less, but it still doesn’t hurt.

Remove all processing from the mix bus before sending. No master bus limiter. No master bus compressor with gain riding. The mastering engineer needs to hear what the mix is actually doing, not what your mix bus is trying to hide.

What About Genre-Specific Mastering Approaches?

Most guides ignore this. We think that’s lazy.

Hip-Hop and Trap

The 808 is the centre of the universe. Everything else orbits it. Use mid-side processing to keep the 808 mono and locked. Low-end imaging above about 120 Hz is fine. Below that, collapse to mono or you’ll lose the punch on systems that sum to mono. A high shelf cut around 10 kHz at -1 to -1.5 dB keeps the hats from screaming, which they will do in this genre if you let them.

Classical and Acoustic

This is where we hate aggressive limiting. Classical dynamics can swing 20+ LU between a pianissimo passage and a full orchestra. Your limiter should be doing 0.5 dB of reduction maximum. You’re not making it loud. You’re making sure nothing clips and the presentation is clean. Target -23 LUFS if the work is headed for broadcast.

Electronic and Synth-Driven Music

Frequency content is fully controlled in the mix, which means mastering EQ problems are almost always mix problems that crept through. Kick and bass relationship at 60-100 Hz is the main battleground. Multiband compression at 2:1 below 100 Hz, with a slow 50ms attack to let the kick transient breathe, is a satisfying fix when it locks in correctly.

We ran a dense techno track through this approach with a 10ms attack and got pumping. Pulled the attack to 45ms and the kick snapped through cleanly while the sub sat underneath it. One parameter. Different record.

Where Does AI Mastering Fit Into This?

AI mastering tools like LANDR, eMastered, and Ozone’s Master Assistant have improved substantially. They’re faster than any human and competent at broad tonal correction on clean, well-mixed material.

The honest assessment: a discernible quality gap remains between AI-only masters and experienced human mastering engineers, specifically in dynamic range preservation and aesthetic judgement. AI tools optimise toward a target loudness profile. They don’t know the track is supposed to breathe at that moment, or that the intentional distortion in the bridge isn’t a problem to fix.

The workflow we’d suggest for smaller budgets: use LANDR or Ozone’s Master Assistant as a reference and diagnostic tool. Let it show you what adjustments it wants to make. Understand why it’s making them. Then make those adjustments yourself, or approve the specific ones you agree with.

The AI catches obvious problems. The human engineer catches everything else.

Hybrid Sessions

iZotope Ozone 11’s Master Assistant will analyse your track and suggest a starting point for the full chain: EQ, dynamics, imager settings. We’ve found it’s a brilliant time-saver for the first 10 minutes of a session. The final 30 minutes of critical listening and micro-adjustment remain human work. That’s not changing soon.

Worth Bookmarking

What Does Professional Mastering Actually Take in Time?

Experienced engineers spend 30-40 minutes per track on a typical commercial release. That time breaks down roughly into: 5-10 minutes critical listening before touching anything, 10-15 minutes chain building and initial adjustment, 10-15 minutes A/B testing, reference comparisons, and final decisions.

The A/B test is non-negotiable. Bypass your entire chain and compare at matched loudness. If the processed version doesn’t win on that comparison, something in the chain is hurting more than it’s helping. This is an annoying lesson to learn for the first time at midnight before a delivery deadline.

Take breaks. Ear fatigue is real, it’s measurable (your sensitivity to 3-5 kHz drops noticeably after 60-90 minutes of loud monitoring), and a decision made at the wrong moment of fatigue costs you the whole master.

Summary

Mastering works when you stay precise and stay humble. Run your signal chain in order: high-pass, broad EQ at +/- 1-2 dB, compression at 2:1 with slow attack, stereo imaging, limiting to -1.0 dBTP true peak. Target -14 LUFS for streaming, -9 to -7 LUFS for commercial pop. Deliver 24-bit WAV files at project sample rate with -3 to -6 dBFS headroom. Use AI tools as diagnostics, not replacements. Always A/B at matched levels. And if you’re pulling 6 dB of gain reduction in your limiter, send the mix back.

Part of our complete Music Mastering Guide: Complete Beginner's Breakdown series.

Frequently Asked Questions

What LUFS should I target when mastering for Spotify?

Spotify normalises playback to -14 LUFS integrated. Mastering louder than that doesn't increase perceived volume on the platform because Spotify turns the track down to match. Deliver at -14 LUFS and you'll preserve more dynamic range without sacrificing anything on playback. If your track is specifically targeting playlists where loudness is a competitive concern, -10 to -9 LUFS is the commercial pop range, but understand that Spotify reduces it on playback regardless.

How much headroom should I leave in my mix before sending it to mastering?

Peaks should land between -3 and -6 dBFS on the mix bus output, measured in peak dBFS rather than RMS. This gives the mastering chain room to apply EQ and compression before the signal hits a limiter. Remove any gain-riding plugins or limiting from your master bus before bouncing. The mastering engineer needs to hear the mix, not your mix bus trying to compensate for it.

Is AI mastering good enough to replace a mastering engineer?

For clean, well-produced mixes on standard commercial genres, AI tools like LANDR and Ozone's Master Assistant produce acceptable results quickly and cheaply. The quality gap compared to a skilled human mastering engineer remains real, specifically in dynamic range preservation and context-sensitive aesthetic choices. Our suggestion is to use AI tools as a starting point and diagnostic reference, not a final delivery solution for anything important.

What's the difference between integrated LUFS and short-term LUFS?

Integrated LUFS measures the average loudness of an entire track from start to finish. Short-term LUFS uses a 3-second sliding window to show loudness in a specific moment. Short-term readings show you whether your chorus is 4 LU louder than your verse, which is useful diagnostic information for compression decisions. Both matter: integrated tells you your platform target, short-term tells you whether your dynamics are working.

Should I use linear phase or minimum phase EQ in mastering?

Linear phase EQ introduces pre-ringing artifacts on sharp cuts but preserves phase relationships across the stereo field. Minimum phase EQ is more natural sounding and causes no pre-ringing, but can shift phase between frequencies. For broad tonal shaping in mastering, minimum phase is fine. For steep corrections or mid-side work where phase accuracy matters, linear phase is the safer choice. FabFilter Pro-Q 3 lets you switch between both on individual bands, which is the approach we'd use on most sessions.

How do I handle a muddy low-mid in a mix I'm mastering?

First, identify whether the problem is at 200-350 Hz (the classic mud zone) or whether it's actually around 100-150 Hz presenting as mud on smaller speakers. Use a narrow-Q dynamic EQ rather than static EQ, so you're only pulling down when the problem frequency actually builds up, rather than cutting it permanently and making the track thin on systems where it's fine. A dynamic EQ band at 250 Hz with a -2 dB reduction triggered at -18 dBFS threshold is a specific starting point. If the mix needs more than -3 dB of correction there, send it back to the mix stage.

What sample rate should I use for mastering?

96kHz is the professional standard for mastering sessions. It gives processing algorithms more resolution to work with and reduces aliasing in saturation and limiting. 44.1kHz is workable if you apply oversampling inside your limiter (most professional limiters including FabFilter Pro-L 2 support 2x to 8x oversampling). If your project was recorded and mixed at 44.1kHz, don't upsample just for mastering. Deliver at the native project rate.

Do I need reference tracks when mastering?

Yes. Always. Reference tracks are not for copying; they're for calibrating your ears and your room against a known quantity. Pick 2-3 commercially released tracks in the same genre, matched to the same LUFS as your master for a fair comparison. The reference should tell you where your tonal balance is wrong, not how loud to make the track. Tools like REFERENCE by Mastering the Mix make A/B referencing inside your DAW straightforward and take about two minutes to set up.