Mixing Fundamentals in 2026

There is no shortage of mixing tips. There is a shortage of mixing order. Almost every fundamentals guide hands you a flat list of twenty things to remember, which is the one format guaranteed not to help, because the value of most mixing advice depends entirely on when you apply it. Cut low end before you have a balance and you will undo it later. Compress before you have ridden the levels and you are compressing a problem instead of a performance.

So this is the same material arranged the way you actually work: six stages, sixteen rules, each one assuming the ones above it are already settled. Every rule also names the specific mistake it prevents, because a fundamental you cannot connect to a consequence is just trivia.

And because some of the most repeated advice in this category is genuinely contested, there is a section further down covering the four points where well-regarded engineers openly contradict each other. Pretending that consensus exists where it does not is how people end up following a rule that is actively hurting their mixes.

Where This List Comes From

Two sources, deliberately different in kind.

What experienced engineers actually teach. I worked through the most-watched fundamentals and workflow material from engineers who publish under their own name and take the reputational risk that comes with it — among them In The Mix, Sage Audio, Andrew Barr, Streaky, Alex Rome, Hardcore Music Studio, Audio University and Waves’ own artist sessions. Where several of them independently arrive at the same rule, that rule is in this list. Where they disagree, that disagreement gets its own section rather than being smoothed over.

What measurement says is actually broken. Advice tells you what engineers think matters; measurement tells you what is going wrong in practice, and the two do not always agree. RoEx analysed 7 million tracks through Mix Check Studio across 180 countries and 30 genres, with the findings presented in a peer-reviewed paper at the Audio Engineering Society’s 157th Convention. Where a rule below has a prevalence figure attached, it comes from there rather than from my impressions.

For the record, this is written by YECK, founder of MixingGPT. There is a section near the end on where an AI assistant helps with this workflow and, more usefully, the five rules here it cannot check at all. Everything before it is tool-agnostic: nothing in the sixteen requires software you do not already own.

The 16 Fundamentals at a Glance

If you want the 30-second version, here it is. Read down the middle column for the rule and the right-hand column for what happens when you skip it. The stages are the point: work them in order and each one holds.

Four of these — rules 1, 3, 4 and 7 — are not about sound at all. They are about arrangement, organisation and timing, and they are the four most likely to be dismissed as filler by someone looking for a plugin recommendation. They also have the highest ratio of result to effort of anything on the list, which is inconvenient but true.

Stage 1 — Before the mix starts

1. Competing Elements Are an Arrangement Problem, Not a Mix Problem

This is the single most valuable thing in this article and the one nobody wants to hear, because it means the answer is not a plugin. When two elements refuse to coexist, the reason is almost never that you chose the wrong EQ. It is that both parts are trying to occupy the same frequency range in the same octave at the same moment, and no amount of carving fixes a collision that the arrangement built in.

How to recognise it: you have spent more than a few minutes EQing two elements against each other and every setting is a trade. One gets clearer and the other gets thin. That trade is the diagnostic. If there were room for both, you would have found it by now.

Why full-range presets make it worse: a modern synth or sample preset is designed to sound impressive on its own, which means it has been given weight in the lows, body in the mids and air on top. Stack three of those and you have three sources each claiming the whole spectrum. As In The Mix puts it, if you want the song to breathe you have to make compromises the preset designer did not.

The fix: change the part, not the processing. Move one element up an octave. Shorten the pad so it stops under the vocal. Thin the voicing of the chords so the bass note is not doubled. Mute a layer entirely and check whether anything was lost. Listen to any record you admire and notice how rarely two things are doing the same job in the same range simultaneously — that is arrangement discipline, not mixing skill, and it is why those records feel effortless.

The honest caveat: when you are mixing someone else’s session, changing the arrangement may not be yours to do. In that case say so rather than burning four hours trying to mix your way out of it. A short note asking whether a layer can be muted is a legitimate mix note.

2. Check Polarity on Anything Layered, Before Any Processing

This takes about fifteen seconds and it is the only item on this list that can invalidate every decision made after it. If two layered sources are pushing in opposite directions, they subtract instead of adding, and nothing downstream can recover what was cancelled before your first plugin saw it.

Where it bites hardest: layered kicks and 808s, stacked snare samples, multi-miked sources, and vocal doubles. The low end is worst affected because long wavelengths make the cancellation broad and obvious in level while staying hard to identify by ear — it does not sound like a phase problem, it just sounds like the kick has no weight.

How to check: zoom in on the two waveforms at a shared transient and look at which way they move first. Both going up, or both going down, is fine. One up and one down is a polarity flip you want. Then flip it and listen: the version with more low end is the correct one, regardless of what the waveform suggested.

Why it survives to release: RoEx measured phase problems in 16 percent of mixes and 15 percent of mastered files. The rate barely improves through mastering, which means it is being caught by neither the person mixing nor the person mastering. Mono compatibility problems follow the same pattern — 17 percent of mixes, still 12 percent after mastering. Fifteen seconds per layered source is a very good trade against those odds. There is a full treatment of the low-end case in how to mix bass in 2026.

3. Consolidate, Label and Commit Before You Open a Single Plugin

Mixing is a long sequence of small judgements, and your capacity for those judgements is finite. Every one you spend hunting for a track, working out what “Audio 14” contains, or deciding again whether you still need three hi-hat layers is a judgement not spent on the mix.

What this looks like in practice: bounce redundant layers down to one region. Name every track something you will recognise in a week. Group the drums, the vocals and the instruments so a level move can be made in one gesture. Print the creative processing you have already committed to instead of leaving twelve plugins open for reconsideration. Alex Rome frames the whole stage as narrowing five layers into two, and the effect is less about CPU than about the number of open questions on screen.

The workflow rule underneath it: Streaky’s version is blunter — do not fear the template. A session template that opens with your buses, sends and naming already in place removes the whole setup stage from every future mix. In The Mix has a related habit worth stealing: if a routine task takes more than about five clicks, stop and find the shortcut, macro or template that removes it. You are paying for the DAW; let it do the clerical work.

Where this goes wrong: committing too early on something you have not decided yet. Print the tuning, the comp, the amp sim and the sound design. Do not print the reverb you are unsure about. The test is whether you would be annoyed to lose the ability to change it.

4. Never Mix on the Day You Wrote It

Writing and mixing use opposite mental modes, and they compete for the same ears. After several hours of writing you are attached to the material, fatigued, and adapted to its exact frequency balance — which is the worst possible starting position for critical listening.

What fatigue actually does: it is desensitisation rather than tiredness. Your auditory system adapts to sustained exposure, so your internal reference for “bright” and “full” drifts over a session. The mix did not change; your yardstick did. This is the mechanism behind the familiar experience of undoing the last hour of work the following morning.

The practical version: finish writing, sleep, open a dedicated mix session the next day. Inside that session, take a break roughly every twenty minutes of concentrated listening. If you have the choice, mix in the morning — Streaky is emphatic on this and the reason is simply that you have not yet spent the day’s listening budget.

Why this outranks most technique: every other rule on this list depends on your ability to hear whether it worked. This is the only one that protects that ability, which makes it a prerequisite rather than a tip.

Stage 2 — Balance, which is most of the mix

5. Everything in Mixing Is Volume — So Use the Fader First

Andrew Barr makes this point more cleanly than anyone: nearly every mixing tool is a volume control wearing a costume. It sounds like a simplification until you notice how much it changes your reflexes.

• EQ is volume for a frequency band. Cutting 300 Hz is turning 300 Hz down.
• Compression is automatic volume, applied conditionally on level.
• Panning is volume distributed between two speakers. Panning right is turning the left channel down.
• Automation is volume you write by hand over time.
• Saturation is the odd one out — it adds harmonics rather than gain, which reads as loudness without a level increase.

What follows from it: reach for the fader before the plugin, every time. Streaky states it as a hard rule — fader before plug-in, always. If a vocal is not cutting through, try the fader before the EQ, because an EQ boost is a narrower and more expensive way to do the same thing. A very large share of what people describe as a mixing problem is a balance they have not tried yet.

The concrete target: get the static fader balance to the point where the song reads correctly with no processing at all. If you cannot get there with faders, adding plugins will not rescue it — you are back at rule 1. If you can, you have finished the majority of the mix before touching an insert.

6. Set the Balance in Mono, Then Open the Stereo Image

This is the closest thing to universal agreement in the whole category, and the reason is not mono compatibility. It is that width lies about level.

The mechanism: a wide source is perceived as louder than a narrow one at the same measured level. Alex Rome calls it fake loudness, and it is why a balance set in stereo tends to leave centred elements — kick, snare, bass, lead vocal, the entire spine of the record — sitting too quietly. Fold to mono and that flattery disappears, leaving only the actual level relationships to judge.

How to set it up: put a mono fold on the master bus and bind it to a key command, because the value comes from switching constantly rather than from working in mono for an hour. Most DAWs have a stock utility that does it. Balance in mono, EQ in mono, then release it and place things across the image.

Then pan outward, not inward: starting narrow and widening produces a cleaner image than starting wide and trying to rein it in. Keep kick, snare, bass and lead vocal centred; everything else is available for placement. That convention exists because those four carry the weight of the record and the centre is the only position guaranteed to survive every playback system.

The counterpoint worth knowing: mono is a diagnostic, not a target. Checking mono is how you confirm you have not broken anything; it is not an argument against width. Mixes that are too cautious and narrow are a real failure mode too, and a common one.

7. Stop Soloing — Judge Every Move in Context

Soloing is a learning tool that has been mistaken for a mixing tool. It is genuinely useful when you are working out what a compressor does. It is actively harmful when you are deciding how loud a snare should be.

Why it costs you time: almost anything sounds better soloed. Add reverb to an isolated vocal and it improves every time. Then the arrangement returns and the decision was wrong, so you make it again. In The Mix identifies this as one of the biggest silent drains on mixing speed — not because the moves are wrong, but because they get made two or three times.

What to do instead: EQ against the element you are competing with. Mixing the bass, keep the kick in. Mixing the vocal, keep the instrumental in, even at a reduced level. Alex Rome demonstrates the same habit when carving piano against bass: the point is not to make the piano sound good, it is to find the setting where both sound full simultaneously, and that setting is invisible in solo.

The secondary benefit: less ear fatigue. Working in context means fewer total repetitions of the same eight bars, which directly protects rule 4.

8. Level-Match Your Reference Track or Do Not Bother Using One

Reference tracks are the highest-leverage habit in mixing and the most commonly botched, because the step everyone skips is the one that makes them work. A mastered commercial track is roughly 10 dB louder than your unmastered mix. Compare them directly and you have learned one thing: mastered music is louder. You already knew that.

The setup that actually works, assembled from how several engineers describe doing it:

• Pick a reference in your genre, from the last couple of years, with a comparable arrangement. A rock reference for a house track teaches you nothing actionable.
• Route it past your master bus, to a separate output. It is already mastered; it must not go through your mix bus processing or your limiter.
• Match loudness before you listen. Attenuate the reference until its integrated loudness matches yours — around −18 LUFS is a convenient meeting point for an unmastered mix. Either a metering plugin or a reference plugin with automatic gain matching will do this.
• Compare the busiest section, aligned to the same point in the bar, and skip incidental moments like a cymbal crash that will distract you.
• Switch fast and repeatedly. The information is in the instant of the switch, not in listening to each for thirty seconds.
• Check it in mono too, and on both headphones and speakers.

The trick worth stealing: put a matching low-pass filter on both your mix and the reference at around 130 Hz and compare only the low end. Isolating the region makes kick and bass level differences obvious in a way that full-range comparison hides. The same works for the top end.

What to listen for, in order: overall balance between the lead element and everything else, low-end weight, and top-end brightness. Those three account for most of the perceived gap between an amateur and a commercial mix. Full method in how to use reference tracks when mixing, and a complementary level-setting approach in balancing mix levels with pink noise.

The objection, answered: referencing is not copying. You are matching technical benchmarks — balance, weight, brightness — not musical content. Producers who avoid references on originality grounds generally adopt them later and wish they had started sooner.

Stage 3 — Tone

9. Think in Fundamentals and Harmonics, Not in Named Frequency Ranges

Every beginner learns the spectrum as a set of labelled zones: sub, bass, low-mid, mid, high-mid, treble, air, with adjectives attached — warm here, muddy there, harsh over here. Sage Audio makes the case that this model is not just simplified but actively obstructive, and it is the most useful reframing in this article.

The problem with ranges: they are too wide to act on and the adjectives are subjective. “Cut some low-mid mud” describes a span of several hundred hertz and gives you no starting frequency, so you sweep and guess.

The replacement model: any single source’s frequency content is four things, not seven zones.

• The fundamental — one frequency, the pitch of the note. A bass playing A1 has a fundamental at 55 Hz.
• Harmonic overtones — whole-number multiples of the fundamental: 110 Hz, 165 Hz, 220 Hz. These are musical by definition, and their amplitude generally falls as you move up.
• Inharmonic overtones — everything performance-related that is not a multiple: the click of the pick, fingers on the fretboard, body resonance, vocal sibilance, snare rattle.
• Non-musical content — hum, ground-loop noise, rumble, handling. Nothing to do with the note.

Why this changes what you do: the guesswork collapses. If the low end of a bass is too heavy, it is heavy at the fundamental or at the second or third harmonic, and you know exactly what those frequencies are for the note being played. If a vocal turns boomy every time the singer hits A2, the culprits are 110 Hz, 220 Hz and possibly 330 Hz — two narrow bells rather than a wide dynamic band and a hopeful sweep. If a vocal is hard to understand, boosting a harmonic near the third formant is a targeted move rather than a blind high-mid lift.

The honest limit: this works cleanly on pitched, largely monophonic sources — bass, lead vocal, a single synth line. On a full drum kit or a dense polyphonic bus, inharmonic content dominates and you are back to listening and trialling. It also does not replace ear training; it just stops you wasting the ear training you have on questions arithmetic could answer.

10. High-Pass With a Reason, Not by Reflex

The most repeated rule in mixing is “low-cut everything except the kick and bass.” It is also the most contested, and both camps have a point, so here is the version that survives scrutiny.

What is true about it: sources genuinely do carry low-frequency content you cannot hear and do not want. Put an analyser on a hi-hat and there is energy well below anything musical in it. Multiply that across thirty channels and you have real energy consuming headroom and clouding the sub region, which is exactly why a mix can measure loud and still feel weak. Cutting it costs you nothing on that hi-hat.

What is wrong about it: the rule gets applied as a number rather than a judgement. A filter at 150 Hz on every channel is not housekeeping, it is removing the fundamental of the low piano notes, the body of the snare, and the bottom of the rhythm guitar. Do it to thirty tracks and the mix gets thin in a way that is hard to trace, because no single filter was audibly wrong.

The resolution, which is rule 9 applied: find the lowest fundamental the source actually plays, and filter below that. Then everything you remove is non-musical by definition, and nothing you remove is part of the performance. For a hi-hat that lands high and the filter is free. For a piano it lands low and the filter is nearly pointless — which is the correct answer, not a failure.

Two refinements: a gentle low shelf is often better than a steep cut, because it thins the region without deleting it, and it sounds smoother. And decide per source whether the low end needs to be mono. Collapsing sub content to mono is near-universal in electronic and hip-hop, and a genuine loss on wide rock and acoustic productions where the low end is part of the image.

The tell that you have gone too far: the mix passes a mono check, measures fine, has plenty of headroom, and sounds small. That combination is almost always accumulated filtering. Related reading in how to fix muddy vocals and the best EQ plugins in 2026.

11. Cut to Make Space, Boost to Add Character

“Cut, do not boost” is stated as dogma often enough to be worth restating properly, because the reasoning matters more than the rule. Cutting and boosting are not moral categories; they solve different problems.

Cut when the problem is competition. Two elements want the same region, so you remove it from the one that needs it less. Andrew Barr’s framing is the clearest: if you want a clear sky you remove clouds, you do not add more sky. Pushing the vocal harder to beat the guitars raises the total energy in the region and makes the whole mix denser; carving the guitars where the vocal lives achieves the same separation and lowers it.

Boost when the problem is character. A snare that needs more crack, a vocal that needs air, a bass that needs more of its second harmonic to read on small speakers — these are additive goals and a cut cannot achieve them. Boosting is fine. Boosting as a substitute for balance is not.

The practical constraint: boosts accumulate into gain, and gain accumulates into the limiter. This is where tone quietly becomes a gain-staging problem, and it is worth watching the master while you work rather than discovering it at the end. Groundwork in the ultimate guide to gain staging.

The bypass test: after any EQ move, bypass it and level-match. If you cannot pick which one is better, remove it. Streaky’s version applies to the whole mix — if something is not adding anything, mute it and see whether anyone would miss it.

Stage 4 — Dynamics

12. You Are Far More Likely to Be Under-Compressing Than Over-Compressing

This is the one place where the popular advice and the measured data point in opposite directions, and the gap is large enough to be worth sitting with.

Almost every tutorial warns against over-compression: do not squash it, do not kill your transients, do not engage the bus compressor too early. Meanwhile RoEx found that 46 percent of mixes are under-compressed — dynamic range too wide, quiet parts too quiet, no cohesion. Nearly half of real mixes have the opposite problem to the one the internet is warning about.

Why the under-compressed case is invisible: it sounds correct when you audition your own track alone. The failure only appears in comparison, next to records that hold together, which is why it survives all the way to release. RoEx also notes it is the problem most often misdiagnosed as a mastering issue — and mastering cannot fix it, because it acts on the summed stereo file rather than the individual elements that need controlling.

How much is normal on a lead vocal: considerably more than most people expect. Hardcore Music Studio walks through EQ, reverb, layering and automation on a raw rock vocal, demonstrates that none of them produce the commercial sound, and then shows that compression does — a compressor on the way in at 6 to 12 dB of gain reduction, a console-style compressor in the mix, and an 1176-style compressor after it working hard. Twenty dB or more of total gain reduction across the chain is ordinary. Joe Barresi describes the same approach on hard rock vocals: several compressors in series, each doing a little, which gives more control and colour than one doing all of it.

Why that much: on a vocal, compression is not only level control. It sharpens the attack of consonants so the delivery reads as more energetic, and it lifts low-level detail — breaths, mouth sounds, room tone — toward the front, which is most of what “close and intimate” actually means. The well-known tell is that on many commercial records the breaths are as loud as the words. That is not an accident.

Where the warning is still right: on the mix bus, and on drums where transients carry the groove. The distinction that resolves it is placement rather than amount — be generous per element, conservative on the summed mix. And In The Mix is worth quoting on the folklore: the widely repeated rule that you should never compress more than 3 to 6 dB is not a professional standard, it is internet gospel. You do what the track needs.

What to check: peak-to-loudness ratio and short-term crest tell you how much dynamic range is left; loudness range tells you whether sections still move. A very low loudness range on a song with a quiet verse means you have flattened it. An unusually high one on a dense modern production usually means nothing is being controlled at all. More in compressing vocals with RVox, 1176 and CLA-2A and parallel compression done properly.

13. Ride the Fader Before You Reach for a Harder Compressor

A compressor reacts to level. It cannot tell the difference between a phrase the singer leaned into deliberately and a phrase that drifted off-mic, so it treats both as the same event. That is why a performance with wide swings has no single correct fader position and no single correct threshold either.

The order that works: even out the obvious level differences first, by clip gain or by writing volume automation, so the compressor receives a signal that is already broadly consistent. Then it can do character work on a stable input instead of spending its whole range fighting the performance. Reverse this and you end up compressing harder and harder to control a handful of moments, which flattens everything else.

What to automate beyond level: effect sends, so a delay throw appears in the gap after a line rather than under it. Compression amount itself, so a sparse opening verse gets less and a dense final chorus gets more, keeping the vocal energy matched to the arrangement. And, sparingly, the mix bus.

Why this is the dividing line: it is the clearest single difference between an amateur and a professional mix. Amateurs set a chain and leave it. Professionals ride things continuously, because the correct level for a vocal in bar 9 is not the correct level in bar 57. Step-by-step in how to automate vocals.

Stage 5 — Space

14. Reverb and Delay Are Sources That Need Mixing Too

Reverb is the widest, most full-range signal you will add to a mix, and it is the one most often inserted, approved in solo, and never revisited. Alex Rome makes the consequence concrete: put an analyser and an imager on a reverb return in isolation and you find energy across the entire spectrum and across the entire stereo field. Do that on eight channels and you have added eight full-range sources nobody mixed.

The symptom: a mix that is cluttered and will not get loud, with no individual element to blame. It is also why heavily reverberant vocals so often bury the track — the wash is competing with the arrangement, not sitting behind it.

Treat the return like a channel:

• Filter it. High-pass the return generously and roll off the top. This does not make the reverb sound worse; it makes it fit.
• Use pre-delay. A short gap between the dry signal and the onset of the tail keeps the word intelligible while still placing it in a space. It is the cheapest clarity available in this stage.
• Time the tail to the track. A tail still ringing into the next bar smears the groove. Shorten it, or gate it, so it resolves in time.
• Consider delay instead. Short delays create depth with far less spectral cost than a long reverb, which is why plenty of engineers reach for them first on vocals.

The technique worth learning: duck the effects with the dry signal. Put a compressor at the end of the reverb and delay buses and sidechain it to the lead vocal, so the wash pulls back while the vocal is sounding and blooms in the gaps. Streaky demonstrates this with roughly 5 to 8 dB of gain reduction, fully wet, no makeup gain. The result is a vocal that stays completely intelligible while carrying far more effect than a static send could support. Big Z uses the same principle on the instrumental instead — a dynamic EQ on the instrumental bus keyed to the vocal, carving 2 to 3 dB of moving space rather than a fixed notch. Both are the same idea: make the space conditional on the thing that needs it.

Where it goes wrong: deciding how much is enough in solo. Reverb always sounds better isolated, which is rule 7 with a different costume. More in vocal reverb plugins and settings.

Stage 6 — Finish

15. The Mix Bus Is Polish, Not Repair — and the Limiter Goes Last

Streaky states this as two consecutive rules and they belong together: get the mix working before you touch the stereo bus, and let the stereo bus be the polish rather than the repair. The temptation runs the other way, because mix bus processing makes an unfinished mix sound momentarily more finished, which is exactly what makes it a trap.

What the mix bus is good at: a small amount of glue compression on a balance that already works, gentle broad tonal shaping, and saturation for cohesion — often a better glue than a compressor, because it adds harmonic density without modulating level.

What it cannot do: fix a balance. It acts on the sum, so it cannot reach one element. If the vocal is buried, no mix bus chain unburies it. If the low end is fighting, no mix bus chain separates it. Detail in inside a professional mix bus chain.

The limiter is last, with two exceptions. A limiter increases level, so anything after it gets driven harder than you intended — which quietly re-introduces distortion you thought you had avoided. The only things that belong after it are metering plugins and headphone-correction monitoring tools, because neither is part of the signal you print.

The measured reason to care: RoEx found clipping in 31 percent of mixes and 57 percent of masters. More than half of finished masters have distortion baked in permanently, and once it is printed it cannot be removed. Over half of the mixes they saw also arrived louder than −17.5 LUFS, leaving the mastering stage nowhere to work. For a mix going to mastering, roughly −18 to −20 LUFS with peaks several dB below full scale is the target. Checklist in how to prepare your mix for mastering.

16. Print It, Walk Away, Judge It Cold on a Second System

There is a large amount of material on mixing technique and very little on mix assessment, which leaves most producers with plenty of moves and no reliable way to tell whether a move helped. This rule is the whole of that missing skill.

The procedure: print a version. Leave it overnight. Play the first thirty seconds on something that is not your mixing setup — a phone speaker, the car, cheap earbuds — and pay attention to your first reaction before you start analysing. That first impression on fresh ears is the most honest data you will get about the mix, and it expires within about a minute of listening.

Why a second system and not better ears: a phone speaker is a genuinely useful instrument. It has almost no low end and a narrow image, so it strips the mix back to whether the balance and the arrangement work. Anything that survives a phone and a car is very unlikely to fall apart anywhere else.

How to know you are done: settle the technical layer objectively so it stops being a matter of opinion — integrated loudness, true peak, clipping count and mono stability should be confirmed facts. Then, if the balance still reads correctly cold on a second system, every remaining question is taste. Taste does not converge, so that is the moment to stop. If you want the full diagnostic version of this question, it is the subject of why does my mix sound wrong.

Four Fundamentals the Experts Openly Disagree About

Everything above has broad agreement behind it. These four do not, and if you have ever followed advice that made your mixes worse, one of them is a likely culprit. In each case both positions come from people who know what they are doing, which means the disagreement is usually about an unstated assumption rather than about the facts.

1. Should you high-pass everything except the kick and bass?

The disagreement: multiple widely followed engineers teach a low cut on every channel except kick and bass as a near-automatic step. Others treat reflexive high-passing as one of the most damaging habits in amateur mixing, and one of the most-watched EQ videos in the category is explicitly a pushback against it.

What both sides are actually saying: the pro-filter camp is describing a busy electronic or hip-hop arrangement with thirty channels, where accumulated inaudible low end is a real headroom problem. The anti-filter camp is describing sources with musical content down there, where a fixed filter frequency removes the performance. Both are correct within their own context and the rule travels badly outside it.

My read: filter with a reason, per source, using the lowest fundamental as the boundary — which is rule 10. Note that even engineers who advocate cutting add caveats: Streaky teaches removing low-end junk from everything but kick and bass and then, thirteen tips later, warns to go easy on it and prefer a shelf to a full cut. That is not a contradiction, it is the nuance getting compressed out of a tip list.

2. How much compression is too much?

The disagreement: the dominant message online is a warning against over-compression. A significant group of working mixers argue the opposite, that most amateur mixes are barely compressed at all and that this is the main thing separating them from records.

What settles it: the measurement. 46 percent of mixes in a 7-million-track dataset were under-compressed. That does not prove over-compression never happens; it proves the warning is aimed at the less common failure. The reason the folklore persists is that over-compression is audible and embarrassing while under-compression is silent — it just sounds slightly unprofessional in a way nobody can name.

My read: the amount is the wrong variable. Be generous per element, conservative on the mix bus, and use gain reduction as a means rather than a target. Anyone quoting a universal dB figure is repeating something they did not measure.

3. Are the loudness wars over, or is your master too loud?

The disagreement: this is the sharpest contradiction in the whole category. In The Mix, after surveying what professional mastering engineers actually deliver, concludes that the loudness wars are over and loud won — that practitioners largely disregard platform recommendations and master for the song. RoEx, from measurement, concludes that 79 percent of masters exceed Spotify’s −14 LUFS and 92 percent exceed Apple Music’s −16, and recommends targeting the platform figure with a −1 dBTP ceiling.

Both are right, about different things. The specification is not a rule and never was — Spotify publishes it as a mastering tip, not a requirement, and mastering engineers who ignore it are making a defensible artistic call. But normalisation is not optional: Spotify adjusts playback to −14 LUFS, which means limiting applied above that target is discarded on playback while the distortion it caused is kept. And 57 percent of masters clipping is not a stylistic choice, it is an accident.

The detail almost nobody mentions: normalisation is not symmetrical. Spotify caps positive gain using true peak headroom — their own worked example is a track at −20 LUFS with a true peak of −5 dBFS being lifted only to −16 LUFS, not to −14. So a very quiet master with low peaks does not simply get turned up to match everything else. That is a real argument against mastering far below the target, and it sits alongside the argument against pushing far above it.

My read: separate the two questions. Loudness is taste, so be as loud as the song wants and audition several versions. Clipping and true peak are not taste — check them, every time, before the file leaves your machine. Detail in how to mix for streaming.

4. Does the gear actually matter?

The disagreement: “it is not the gear, it is the ear” is the standard encouragement. In The Mix pushes back hard, pointing out that the phrase is usually delivered by someone sitting in a treated room in front of expensive monitors, and that no amount of skill compensates for a monitoring path that hides the problem.

The resolution is a category split. Source gear matters far less than people think: Finneas recorded the vocals that launched Billie Eilish’s career on an Audio-Technica AT2020 in his bedroom, moving to a Neumann only later — a fact he has discussed openly in interviews about that room. Monitoring gear matters most, because a decision requires hearing the thing you are deciding about.

My read: spend on monitoring and on the room before anything else, and treat “it is all about the ears” as half true — your ears can only be as good as what feeds them. But the practical answer is not a shopping list. It is one monitoring path you have learned properly, checked against a phone speaker and a car, which is the cheapest reliable setup that exists.

Which Fundamental Are You Most Likely to Be Breaking?

The rules above are ordered by workflow. This table is ordered by how often the failure actually shows up in measurement, which is a different question and a more uncomfortable one. All figures are from RoEx’s 7 million tracks.

Two things fall out of that. The most prevalent failures cluster in the last stage — loudness, headroom and dynamics — which is the least glamorous part of mixing and the part most often rushed at the end of a session. And phase and mono problems persist at nearly the same rate in mastered files as in mixes, which means nobody in the chain is catching them.

RoEx also found the tonal patterns are genre-specific rather than universal: electronic genres consistently showed exaggerated bass, acoustic and folk the opposite, and hip-hop the highest rates of clipping. That is the strongest available argument for rule 8. A target that is right for one genre is wrong for another, and a level-matched reference in your own genre encodes that automatically.

If You Only Have an Hour

Sixteen rules is a lot to hold in your head on a deadline. Compressed to the moves with the best return per minute:

• Five minutes. Zero every fader, fold to mono, and rebuild the static balance from the most important element outward. No plugins.
• Ten minutes. Drop in a level-matched reference and fix the three biggest differences you hear — usually lead level, low-end weight, top-end brightness.
• Fifteen minutes. Ride the lead vocal or lead instrument so it is consistent line to line, then compress it properly.
• Ten minutes. Filter the effect returns and shorten anything ringing past the bar.
• Ten minutes. Open the stereo image back up, then re-check mono to confirm nothing collapsed.
• Five minutes. Check integrated loudness, true peak and clipping, then pull the level back until you have headroom.
• Five minutes. Print it and play the first thirty seconds on your phone.

If you only do one thing from this entire article: check integrated loudness and true peak before you send anything anywhere. Those two numbers catch the most prevalent problems in a dataset of 7 million tracks, and both take under a minute.

Where an AI Assistant Fits Into This Workflow — and Where It Does Not

This is the section where I have a commercial interest, so here is the mechanism rather than the pitch, including the rules it cannot help with at all. MixingGPT is a plugin that loads in Logic Pro, Ableton Live, FL Studio, Studio One, Cubase, Nuendo, Reaper, Bitwig, and GarageBand as AU or VST3 on Mac and PC, so the conversation happens in the session rather than in a browser tab.

Which of the sixteen it can actually check. Drop an audio file into the chat and the plugin computes the DSP facts offline from your samples before anything else happens: BS.1770 integrated loudness, true peak, RMS, stereo correlation, per-band spectral balance, harshness and sibilance indices, presence-resonance, dynamics and punch via peak-to-loudness ratio and short-term crest, and clipping counts. Those numbers settle rules 6, 11, 12, 15 and 16 as facts rather than opinions — whether the mix survives mono, whether headroom is gone, whether dynamics are controlled, whether anything is clipping. Longer uploads are split into time-sections with measured timestamps, so a problem that only appears in the last chorus can be pointed at directly.

Tonal balance is judged against genre targets and returned per band as OK, LOW or HIGH. That matters because of exactly the pattern RoEx found: the same low-mid reading that is correct for a dense electronic mix is a problem for a sparse acoustic one. A verdict without a genre reference is not much of a verdict.

The five rules no analysis can verify. This is the honest boundary, and it happens to include the four highest-leverage items on the list:

• Rule 1, arrangement. A stereo mixdown cannot tell you which layer to mute, because it cannot separate them. It can tell you the low-mids are crowded; it cannot tell you the piano is the reason.
• Rule 2, polarity. Cancellation between two layers is invisible once they are summed. The evidence was destroyed before the file existed.
• Rule 3, organisation, and rule 4, timing. These live in your session and your calendar. No amount of DSP reaches them.
• Rule 7, working in context. A finished mixdown looks identical whether you made every decision in solo or none of them.

Other real limitations. It is built to say “I cannot tell that from a stereo file” rather than guess, which is correct behaviour and genuinely less satisfying than a fabricated answer. The Mix Readiness Score it reports out of 10 is an engineer judgement, not a measurement, and it is labelled that way — anything advertising an objective mix score is overclaiming. Outside the measured values it is given, it is prohibited from inventing specific frequencies, dB amounts, ratios or millisecond times, so recommendations arrive as moves to A/B rather than settings to dial in. Audio analysis is in beta: MP3 and WAV only, up to 50 MB, with 4 to 5 minutes the accurate sweet spot. Pro Tools is not supported, because there is no AAX build. It needs an internet connection.

What it costs. Audio bills by decoded duration at 4 credits per started minute, so a 5-minute full critique is 20 credits and a 30-second stem check is 4. Free includes 10 credits a month shared across chat, images and audio. Starter is $9 for 75 credits (about 18 minutes of audio), Pro is $19 for 200 credits (about 10 full critiques), Studio is $49 for 600 credits (about 30). Nothing rolls over. Full breakdown on the pricing page.

When something else is the better pick. If you want a free one-off technical readout with no account, RoEx’s Mix Check Studio does that well. If you want a tool to render a finished mix from your stems rather than help you make the decisions in this article yourself, this is the wrong category entirely — the comparison of 12 AI mixing plugins covers the cloud services that do that. And if your obstacle is rule 4, nothing here helps; go to bed.

Frequently Asked Questions

Related Deep Dives in This Series

This guide covers the sixteen fundamentals at workflow level. Each guide below goes deeper on one stage of it.