The one-sentence version
Mixing is the work of balancing the individual parts of a song — kick, bass, vocal, guitars — so they sit together. You are working with many tracks at once, and you can change any of them.
Mastering is the work of preparing the finished stereo mix for release. You have one file, two channels, and the decisions are about the whole thing at once: loudness, tonal balance, peak safety.
The clearest test: if the problem can only be fixed by turning something up or down inside the song, it is a mixing problem. If it applies to the whole track equally, it is a mastering problem.
What a mix engineer decides
Level and panning between instruments. Whether the snare cuts through. Whether the bass and kick are fighting for the same frequency range. Reverb and delay sends, automation, whether the second chorus should be bigger than the first.
These decisions need access to the separate tracks. Once the mix is bounced to a stereo file, they are locked in. That is why the mix has to be right before mastering starts — not because mastering is fragile, but because there is nothing left to separate.
What a mastering engineer decides
The release loudness, measured in LUFS. Whether the track needs a shelf at the top or a cut in the low mids. How much the limiter should work, and where the true-peak ceiling sits. On an album, how track three sits against track four.
All of that operates on the stereo file. Broad, gentle moves — a decibel here, half a decibel there — because every change affects everything at once.
Where the line actually sits
The boundary is not as clean as textbooks suggest. A mastering engineer can pull 200 Hz down by a decibel; a mix engineer can put a limiter on the master bus. The distinction that matters in practice is access: mixing has the individual tracks, mastering has one file.
This matters for a practical reason. If your master sounds harsh and the harshness is coming from one cymbal, mastering can only reduce that whole frequency range — which also dulls the vocal and the guitars. Going back to the mix and fixing the cymbal costs less.
How to know which one you need
Your track needs mixing work if: instruments mask each other, the vocal disappears in the chorus, the low end is muddy in a way that changes between sections, or something sounds wrong only in one part of the song.
Your track needs mastering if: the mix sounds right but quiet compared to commercial releases, the tonal balance is close but not quite there, or you need it to meet a platform's loudness target before release.
Muddiness is the most common case people bring to mastering that mastering cannot solve. If the mud moves around during the song, it is a mix problem.
What to hand over
Export a stereo WAV at the session's native sample rate — no conversion, no dither if you are staying at 24-bit or higher. Leave the master bus limiter off, or at least back it off so peaks land around −3 to −6 dBFS.
That headroom is not superstition. A mix that already peaks at 0 dBFS gives the mastering chain nothing to work with: every move has to be a reduction, and inter-sample peaks are already past the ceiling before anything starts.
Do not normalise. Do not bounce to MP3 and back. If the mix has automation that only exists in the master bus limiter, print it before you remove the limiter, so what you hand over is what you meant.
Doing both yourself
Most independent producers mix and master their own work, and that is a reasonable way to release music. The difficulty is not technical — it is that you have heard the song two hundred times and your ears have adapted to its problems.
Two things help. Leave time between finishing the mix and starting the master; a day is enough to hear it fresh. And use reference tracks — commercial releases in the same genre, level-matched, played alongside yours.
Measurement helps too, for a specific reason: louder reliably sounds better, which means your ears will tell you the master improved when all it did was get louder. A LUFS meter is the check on that bias.
What mastering cannot fix
A poor arrangement. A vocal recorded through a bad preamp. Phase problems between two mics on the same source. A kick and bass that occupy the same 60 Hz and were never separated.
Mastering makes a good mix release-ready. It does not rescue a mix that was not working — and a mastering chain that promises otherwise is selling loudness, not quality. If a master sounds better only because it is louder, that is worth checking.
Frequently asked questions
Mixing balances the individual tracks of a song against each other — you have access to every instrument separately. Mastering prepares the finished stereo mix for release, working on the whole file at once: loudness, tonal balance and true-peak safety.
Usually yes, for one practical reason: a mix that sounds right in your room is not necessarily at the loudness a platform expects. Spotify and YouTube normalise playback around −14 LUFS, Apple Music around −16. Mastering sets that level deliberately instead of leaving it to chance.
No. Mastering works on the whole stereo file, so it cannot separate a harsh cymbal from the vocal that sits in the same frequency range. If the problem changes across the song, or comes from one element, it is a mix fix — and going back is cheaper than trying to work around it.
For singles, demos and most independent releases, yes. The obstacle is not technical skill but familiarity — after hearing a song many times you stop noticing its problems. Leave a day between mixing and mastering, use reference tracks, and trust the meter over your ears on loudness.
Peaks around −3 to −6 dBFS is a comfortable range. What matters more is that you leave the master bus limiter off: a mix already squashed to 0 dBFS gives the mastering chain nothing to work with, and inter-sample peaks are past the ceiling before anything begins.
Mixing first, always. Mastering operates on the finished stereo bounce, so any change to the mix means re-mastering. Finish the mix, sit with it for a day, then master.
Loudness is part of it, but the measurable part is the whole job: hitting a platform target, keeping true peak under the ceiling so streaming codecs do not clip, and correcting broad tonal imbalances. A master that is only louder is not a master — it is a gain change.