The one rule that makes every other check work
Human ears consistently rate the louder of two otherwise identical tracks as "better" — clearer, punchier, more exciting. It's the oldest bias in audio, and it's why an honest comparison must be level-matched: turn the louder version down until both feel equally loud, then compare (mastering blogs have preached this for years — see Mastering The Mix on level matching). If your mastered file only wins when it's louder, it hasn't won yet.
With that rule in place, run these five checks.
Check 1: The level-matched A/B
Play the original and the master back to back at matched volume and ask one question: did anything get clearer? Good mastering shows up as better-defined low end, vocals that sit more confidently, sections that connect more smoothly. If the level-matched master sounds identical to the mix, that's not failure — subtle source material sometimes needs little. If it sounds worse — smeared, squashed, gritty — trust that verdict.
Check 2: The loudest 30 seconds
Skip to the biggest moment of the track — the final chorus, the drop — and listen for the two signatures of over-processing: pumping (the whole track audibly "breathing" or ducking with the kick or vocal) and distortion (a fizzy, crackling edge on what should be clean peaks, especially cymbals and loud vocals). These artifacts live in the loud sections, and streaming encoders can make marginal ones worse. Thirty focused seconds here beats three casual full playthroughs.
Check 3: The quietest 30 seconds
Now do the opposite: the intro, a breakdown, the fade-out. Raising loudness raises everything, including what used to be inaudible — room noise, generation artifacts, hiss between phrases. If the quiet parts of your master now have an audible noise floor that the original didn't, the processing was too aggressive for the material.
Check 4: The small-speaker test
Play the master on the worst speakers you own: your phone, laptop speakers, cheap earbuds. Most of your audience lives there, not on studio monitors. A master that "translates" keeps its vocal intelligibility and its groove on a phone speaker. If the hook disappears or the low end turns to mud on small speakers, listeners will hear that version — not the one on your good headphones.
Check 5: The reference sandwich
Make a three-track playlist: a commercial release in your genre, then your master, then the reference again — and level-match here too, since commercial tracks are loud. You're not trying to beat a major-label release; you're checking that your track doesn't sound like it walked in from a different room. Noticeably duller, thinner, or more brittle than the reference is actionable feedback. "A bit different in character" is just identity.
What about the numbers?
Meters help, but they confirm rather than decide. For context: Spotify normalizes playback to −14 LUFS by default and recommends mastering around that level with true peaks under −1 dBTP (Spotify's official guidance). Two things matter about those numbers. First, hitting them doesn't make a master good — a track can measure perfectly and still pump and fizz. Second, missing them slightly doesn't make it bad. If the five listening checks pass, the numbers are almost always fine; if the checks fail, no number rescues the file.
If the master fails a check
You have three honest options. Re-master with different settings or a different tool — automatic tools have opinions, and another pass may suit the track better. Fix the source — mastering amplifies what's there, so a harsh or noisy generation is better re-generated than repaired (this is the most common fix for AI-generated tracks). Or release the original — an un-mastered mix that sounds right beats a mastered file that doesn't.
Where Tonekai fits
We build Tonekai, a local AI mastering app for macOS, and this checklist is exactly why its A/B button exists: after processing, you can switch between the original and the mastered version before exporting, so check 1 happens inside the app instead of across two files in a media player. Processing is on-device (nothing uploads), it's Apple Silicon native, and it runs on macOS 14 or later. The app is free to try, with a one-time Pro Unlock — no subscription — for WAV/AAC export and unlimited mastering.
Download Tonekai on the Mac App Store — free to try.
Frequently asked questions
Do I need a LUFS meter? For releasing your own music: no. The checks above catch the problems that matter. A meter becomes useful when a platform or client gives you a specific delivery spec.
How many times should I listen before releasing? One honest pass of the five checks — roughly ten minutes — catches most problems. Fresh ears the next morning catch the rest. Fifty anxious replays mostly catch nothing new and make everything sound wrong.
My master passes every check but still doesn't feel "professional." Why? Usually the gap is in the mix or the arrangement, not the master. Mastering polishes; it doesn't re-record. Compare your original against the reference track at matched volume — if the gap is already there, it's not a mastering problem.
The bottom line
Judging a master is a skill, but it's a listening skill, not a technical one: match levels, interrogate the loudest and quietest moments, check small speakers, and sandwich your track between references. Ten minutes, no plugins, no meters — and you'll know more about your master than any number can tell you.