Finishing and references

You've been working on a mix for days. It sounds good. Then you play a song you love straight after it, and your heart sinks: theirs is bigger, brighter, punchier. Yours sounds small and dull.

Before you start piling on plugins, try this. Here are two finished songs. Switch between them and decide which mix sounds better.

Now the same two songs, matched to the same loudness. Same question.

In the first player, Song B had a 4 dB head start, and it probably won easily. Matched, the contest gets much more interesting. Some of the "bigger, brighter" disappears, and what's left is specific: one has more low end, one has a drier vocal, one has drums that jump out further.

A reference track can turn "mine sounds worse" into a short list of differences you can actually do something about. You can learn from it without copying it.

Why your mix always loses the first round

Released songs are mastered: limited and compressed to sit loud and dense. Your mix, sensibly, isn't. Play them back to back and the reference wins on level alone, for exactly the reason you met in Listening and comparing: louder sounds fuller and brighter.

So the first rule of referencing: turn the reference down to match your mix, never your mix up to match the reference. It keeps your mix honest and your ears rested, and it shows you what's really different.

Pick a reference for one reason

A good reference shares something with your song: similar instruments, similar energy, similar style. A huge rock chorus is a strange target for a quiet acoustic verse. That doesn't make it a bad song; it just answers a different question.

Decide what you want to learn before you press play:

• How loud does the vocal sit against the band?
• How much low end is there, and is it the kick or the bass carrying it?
• How wet is the space? How wide?
• How hard do the drums hit?

And compare like with like: their chorus against your chorus, their verse against your verse. Their loudest ten seconds against your quietest intro tells you nothing.

Two kinds of loudness

When you match songs, it matters what you measure.

• Integrated loudness is the average over the whole song, from start to finish. It's the number streaming services use when they turn tracks up or down.
• Short-term loudness is a moving three-second average. It tells you how loud this section is.

Match two songs on their whole-song numbers and their choruses may still be several units apart. Match them on the chorus and the verses may be off. For section-by-section listening, match the sections. For a delivery check, reset the meter and play the whole song from start to end in one go. A number from half a song is half an answer.

Three observations, one fix

Now listen to your mix and the reference, matched, on the same section. Listen three times, for three different things: the focus (can you follow the voice or lead as easily?), the low end, then the space and tone.

Write down at most three differences, and make them specific. "My vocal's last words disappear in the chorus" is something you can fix. "Theirs sounds more professional" isn't.

Then ask, honestly, whether each difference is actually a problem. Your song might be meant to be darker, or rawer, or drier. A reference can show you a choice without proving your choice is wrong.

Fix one thing. Then compare again, at matched level, on the same section. If the improvement survives, keep it. If it vanishes, it was probably the level fooling you again.

Then check a different section. A brighter master EQ might sparkle in the chorus and make the verse harsh. A limiter that tames one peak might flatten a drum fill you loved.

Loudness targets come from where the song is going

There's no single "correct" loudness. Most streaming services turn loud songs down by default so everything plays at a similar level, which means a squashed master gains nothing there and loses its punch. A club DJ, a radio station, a film or a game may all want something different. Check the current requirements of wherever your song is going, because they change.

What you can control is the trade. More limiting buys loudness and costs dynamics: the drums hit less hard, the chorus lifts less from the verse. Decide it by listening at matched level, not by chasing a number. Limiting and loudness goes deep on this.

Keep an eye on true peak too: it catches peaks that land between samples, which can distort when the song is converted or encoded later. Leave a little room below 0 dBFS.

Check it on other speakers

Play the mix on a few systems you know well: your monitors, headphones, a phone, a laptop, maybe a car. Play the reference on each one too. That shows you what's the speaker and what's the mix.

You're not trying to make it sound identical everywhere; that's impossible. You're looking for failures: the bass line vanishes, the vocal hides, the hi-hats hurt. If three systems point to the same problem, fix it. If one unfamiliar system sounds odd, play the reference there before you blame your mix.

Knowing when to stop

A mix is never perfect. It's finished when:

• It tells the story you meant it to tell, and you could say who leads each section.
• The delivery checks pass: loudness and peaks where the destination needs them.
• Your last few changes don't survive a fair comparison. When tweaks only sound better until you match the level, you've reached the end.

Before the last pass, write down what's left ("automate the vocal in the bridge, clean up the ending, final loudness check"). Do those, listen once through without touching anything, and export. Save the session and the notes, so a future revision starts from what you knew, not what you remember.

Same idea, different place: you can study how a reference balances its bass against its vocal by pulling it apart with Stem separation. Just remember the separated parts are a model's estimate, not the original tracks, so treat what you hear as a clue, not a fact.

Make it yours

Pick one reference song you know inside out, in a style close to yours, and one of your own mixes. Load both into your own session and turn the reference down until their choruses sound equally loud.

Write three specific differences. Fix only the one that matters most. Compare again at matched level, then check the verse.

Save three things together: your notes, the before and after exports, and one sentence on why this version is done, or what one thing is left. That sentence is how you'll know, next time, when to stop.

Quick check

1. After matching loudness, your mix still sounds smaller than the reference in the chorus, mostly because the reference vocal is drier and more upfront. Do you need a louder master? No. That's a balance and space difference. Try less reverb or a slightly louder vocal in the chorus, then compare again at matched level.
2. You match your mix to a reference using the integrated loudness of each whole song. Their chorus still sounds louder than yours. Is the meter wrong? Not necessarily. Their song may have quieter verses and a bigger lift into the chorus, which the whole-song average hides. Compare the choruses with short-term loudness, or by ear.
3. Every change you make in the last hour sounds better at first, but when you level-match against the earlier version, you can't tell them apart. What does that tell you? You're probably done. The improvements were mostly level. Export, and give your ears a rest.

Next steps

• Explanation: Level and loudness. Read the difference between peaks and perceived loudness before choosing your final level.
• Go deeper: Drum programming. At the same loudness, the reference's drums still hit harder. Is that the mix, or the way the beat was programmed?