Train six things, in this order: loudness, frequency, dynamics, space, stereo, and diagnosis. Most people train frequency identification, get reasonably good at it, and stop — which is exactly where progress stalls. The order matters because each layer sits on the one under it: you can’t judge whether a compressor helped if a half-decibel level difference is quietly telling you it did.
What to train, in order
Six skills, each one leaning on the previous.
| # | Skill | What you’re actually learning | Sign you can move on |
|---|---|---|---|
| 1 | Loudness | Hearing a 1 dB difference as louder, not as better | You match two clips by level before judging them, automatically |
| 2 | Frequency | Naming a boosted band, then a cut, then a narrow resonance | You can name an octave-wide band on unfamiliar material |
| 3 | Dynamics | Spotting that compression is happening at all, then attack and release | You hear a fast attack kill a transient without looking at a meter |
| 4 | Space | Reverb type and size, then wet amount, then predelay | You can tell a plate from a hall on a vocal, blind |
| 5 | Stereo | Width, phase, and what survives the mono button | You predict the mono collapse before you press it |
| 6 | Diagnosis | Hearing a problem when nobody told you there is one | You open a mix and know your first move |
None of this is a law. It’s an order that stops each skill from being sabotaged by the one you skipped.
Why loudness comes first
Because louder sounds better, and that single fact ruins more mixing decisions than bad taste ever has. A boost of a decibel or two makes a track sound fuller, closer, more "present". It is none of those things. It is louder.
Until you hear a small level change as a level change, every other comparison is contaminated. You’ll pick the plugin with more output gain. You’ll keep the EQ move that added 1.5 dB of broadband energy and call it warmth.
So the first drill is boring: two clips, one slightly louder, which one and by how much. Get to where 1 dB is obvious and 0.5 dB is a coin flip you sometimes win. That’s enough. After that you level-match before you judge, because you know what happens if you don’t.
Frequency: the big one, boosts before cuts
Frequency identification is what everyone means by "ear training", and it deserves the biggest share of your time — just not all of it.
Train boosts before cuts. A boost adds something that wasn’t there, so your ear has an event to latch onto. A cut takes something away, which is a much quieter kind of change, and beginners routinely miss a 6 dB cut that they’d catch instantly as a boost. Same with width: start wide (a whole octave, generous gain), then narrow the Q, then drop the gain.
The useful landmarks are the ordinary ones every engineer uses as shorthand: rumble and mud low down, boxiness in the low mids, honk and nasal quality in the mids, harshness in the upper mids, sibilance above that, air at the top. Treat those as rough neighbourhoods, not addresses. Different sources put the same word in different places, and a "boxy" kick and a "boxy" vocal aren’t at the same frequency. We go through those bands and how to learn them in our guide to identifying frequencies by ear.
Then add resonance hunting: one narrow peak, somewhere, find it. That’s the one that turns into real work, because most of what you fix in a mix is a resonance, not a band.
Dynamics: first "is it compressed", then attack and release
Start by learning to hear that compression is happening at all — before you try to name the attack time.
This is the step people skip hardest. They read about attack and release, set up a drum loop, and try to hear 10 ms versus 30 ms while they still can’t tell a compressed vocal from an uncompressed one at matched loudness. Same mistake as training cuts before boosts: too small a change, too early.
Order that works: is it compressed at all → roughly how much (ratio and threshold together, as "how squashed") → attack, because it changes the transient and transients are loud events → release, which is the subtlest of the four and shows up as pumping and breathing rather than as a shape. If you want the listening cues for each of those, we wrote them up separately.
Space and stereo: type before amount, width before mono
For reverb, learn type and size before predelay. Hall, plate, room, spring — those are big character differences, and they train the part of your ear that later notices a 20 ms predelay pushing a vocal forward.
Then wet amount, which is really a loudness skill in a costume. Then predelay and tail length, which are genuinely hard and worth leaving until the rest is solid.
For stereo, train width first, then mono compatibility, then phase and comb filtering. Mono compatibility is the one that saves you: a wide synth that vanishes on a phone speaker is a problem you want to hear before you print. Comb filtering — that hollow, flangey sound from two near-identical signals slightly offset — is the last to click, and then you hear it everywhere.
Diagnosis: the skill nobody trains
The last chapter is hearing a problem when nobody told you a problem exists.
Every quiz format has the same tell: it announces that something was done. "Which band was boosted?" already told you a band was boosted. Real mixing doesn’t do that. You get a rough mix and silence.
So the final drill is open-ended: play a clip, decide whether it needs anything, and if so what you’d reach for. Sometimes the honest answer is nothing. Learning to say "nothing" is part of it.
The discipline matters more than the tool
Three habits do more for your ears than any app choice, ours included.
- Level-match everything. Before/after, plugin A versus plugin B, your mix versus a reference. If the levels aren’t matched, you are not comparing what you think you’re comparing.
- One pair of headphones. Not the best pair — the same pair. You are learning what a frequency sounds like through a specific system, and swapping systems every week resets that. Add a second reference later, once the first one is in your bones.
- Short and daily. Ten minutes a day beats an hour on Sunday. Ear training is fatiguing in a way that hides itself: after twenty minutes you’re still answering, just worse.
One more: train on material that resembles what you actually make. Pink noise drills are efficient and transfer poorly. If you produce electronic music, drill on synths and drums — bounce loops from a free Serum 2 preset pack or from whatever you already have open. It’s also why the audio in our own trainer is stems and mixes from our sample packs rather than tones.
How to check yourself honestly
Use a blind ABX test, because your opinion about your own ears is worthless.
ABX gives you A, B, and a hidden X that is one of them, and asks which. Do it enough times and you get a number that doesn’t care how confident you felt. It teaches nothing on its own — it tells you whether a difference is real to you, or whether you’ve been agreeing with your eyes. Run it on things you already believe: two EQ curves you think are different, a compressor at 2:1 versus bypass, the saturator setting you always use.
Where a trainer fits
A trainer’s job is reps and honest feedback. It doesn’t teach you to mix, and any tool that implies otherwise is overselling.
Our ear training game sequences roughly the order above — full disclosure, it’s ours, so weigh that accordingly. Ten chapters, 500 levels, starting with gain and loudness, then EQ, compression, saturation, reverb, stereo and phase, ending with "what would you fix" diagnosis levels. It’s free, browser-based, and the first ten levels don’t need an account. Answer formats include ABX and a fix-it mode where you match a reference with two knobs, which is closer to real work than multiple choice. Limits worth knowing: it’s a game, not a curriculum, there’s no video teaching, and browser-only means no offline app.
If you want to compare it against the other options honestly, we keep a running comparison of SoundGym and its alternatives, checked August 2026. Several of them are free, and one or two do specific things better than we do.
For a bigger pool of source material, our Serum 2 preset bundles are the same audio our exercises are built from — real stems and full mixes, not test tones.
FAQ
What should I train first for mixing?
Loudness. Learn to hear a 1 dB level difference as a level difference, not as an improvement, because small volume changes make anything sound better and will corrupt every comparison you make afterwards. Once level-matching is a habit, move to frequency identification, which is the biggest single skill.
How much ear training should I do per day?
Ten to fifteen minutes, daily, beats a long weekly session. Focused listening tires you out faster than it feels like it does, and past roughly twenty minutes your accuracy drops while your confidence doesn’t. Short and consistent also keeps it a habit rather than a chore you skip.
Should I train frequencies with pink noise or with music?
Both, for different reasons. Pink noise is efficient because a boost is unmistakable, so it’s a good way to first attach a name to a band. But it transfers poorly on its own — drill on real material, ideally the kind you produce, so you learn what 300 Hz sounds like on a kick, a vocal and a pad rather than in the abstract.
Do I need expensive headphones for ear training?
No. Consistency matters far more than quality: you’re learning what frequencies sound like through one specific system, and switching systems restarts that learning. Use whatever you already mix on, keep using it, and add a second reference later once the first is familiar.
Is ear training useful if I only make electronic music?
Yes, and the mixing-focused skills matter more than musical ones. Interval and chord training is musicianship, not mixing — useful, but a different subject. What pays off for producers is frequency identification, spotting compression, reverb character, stereo width and mono compatibility, because those are the decisions you make every session.
How do I know whether a difference I hear is real?
Test it blind with ABX at matched level. You listen to A and B, then to a hidden X that is one of them, and say which — repeated enough times to be more than chance. It won’t teach you anything by itself, but it will tell you honestly whether you can hear a given difference, which is the only way to stop fooling yourself.
I run Monosounds.studio and make the packs myself. Ten-plus years of producing, 97+ releases, 16,000+ downloads. Every Serum 2 preset we ship passes a CPU check on a mid-range laptop and gets used in a real track before release. If it does not survive a mix, it does not go in the pack.
More about Maxim & the studio →
