Using reference tracks properly: level matching, loudness meters and choosing references

Every mixing guide tells you to use a reference track. Fewer explain why so many producers do it and still end up with a mix that is harsh, squashed or thin. The usual reason is simple: the comparison isn't fair. A released track has been mastered, so it plays louder than your mix, and in our experience the louder of two versions usually sounds better at first listen. Switch between them unmatched and you will reach for EQ and compression to close a gap that is really just volume.

This guide covers the three things that make referencing work: matching level properly, reading a loudness meter, and choosing what to compare against. If you are still fixing the basics of a home mix, start with our bedroom mix mistakes guide, where "mixing without a reference" is number ten.

Step 1: get your static mix in place first

Don't load references on an empty session. iZotope's guide to referencing, written by mixing engineer Nick Messitte, says to do the level matching after you gain stage and set your level balance. In other words, get a rough static mix, with faders and pans set and no heavy bus processing, and only then bring the reference in. Otherwise you will be matching the reference to a mix that is about to change anyway.

Practical set-up:

• Put the reference on its own track, routed so it bypasses your master-bus processing. If it runs through your bus compressor or limiter, you are no longer hearing the release as it was made.
• Put a loudness meter where it reads both: on the master bus with the reference track switched in and out, or on each track separately.
• Use a lossless file where you can. Messitte's advice when a client sends a reference is to get a WAV, because it is "always better than the lossy version".

Step 2: match levels with a loudness meter, not a peak meter

Peak meters tell you how high the tallest sample goes. They don't tell you how loud something sounds, and a heavily limited master can peak at the same place as your mix while sounding far louder. Use a loudness meter instead.

Modern loudness meters follow ITU-R BS.1770, the international recommendation for measuring programme loudness and true-peak level; the version in force is BS.1770-5, from November 2023. Readings are shown in LUFS (loudness units relative to full scale). The European Broadcasting Union's Tech 3341 defines the time windows most meters display:

For referencing, short-term is the practical choice. Play the loudest section of your mix, usually a chorus or drop, and note the short-term reading. Play the matching section of the reference and turn the reference's fader down until its short-term reading is the same. iZotope's own walkthrough shows exactly this: a reference pulled down by 4.1 dB to read -17 LUFS short-term.

Two cautions:

• Match like with like. Comparing your verse to their chorus gives you a meaningless number. Pick equivalent sections.
• Trust your ears last. After matching on the meter, flip back and forth. If one still feels louder, nudge it by ear until they feel equal. The meter gets you close; the goal is equal perceived loudness.

Step 3: listen at a sensible level, and switch quickly

Your monitoring volume changes what you hear. iZotope's guide notes that if you monitor too quietly, the midrange feels more prominent and the bass and treble weaker. It cites 85 dB SPL as a commonly recommended monitoring level, while adding that 85 dB is not always right for every room and smaller rooms may need less. The useful habit is consistency: pick a level, mark the knob, and do your reference comparisons there.

Switch quickly. Messitte suggests taking a ten-minute silence break after an hour or two of mixing, then playing about five seconds of the reference and switching straight to five seconds of your mix. Short switches catch differences before your ears adjust to them.

What to compare, one thing at a time:

1. Vocal or lead level against the backing.
2. Low end: how loud the kick and bass sit, and which one leads.
3. Overall brightness.
4. Stereo width, and what still holds up in mono.
5. Dynamics: how much the chorus lifts over the verse.

Step 4: choose references on purpose

A reference is a measuring stick, so pick one that measures the right thing.

• Same job, not just same genre. If your track is a sparse vocal ballad, a dense club record from the same genre tag tells you little. Choose something with a similar arrangement, tempo and lead instrument.
• Two or three, not ten. One reference can mislead you into copying it. A few that agree with each other show you the range of what works. Rotating through a dozen just adds noise.
• Tracks you actually love. Messitte's advice is to choose references "that move you the most", because they are for inspiration as well as matching.
• Your client's reference, if there is one. When an artist hands you a track they love, Messitte treats it as your first genre guide. Get the lossless file, and ask what they like about it. "The drums" and "the vocal sound" lead to very different mixes.
• Good sources. Where you can, use a lossless download or the original WAV rather than a low-bitrate stream or a ripped video, in line with Messitte's WAV advice above.

Tools can help. Matching plug-ins and tonal-balance meters plot your mix's frequency spread against a reference. They are useful for spotting a gap, such as a missing low end, but they are reading curves, not hearing the song. Use them to check what your ears suspect, not to tell your ears what to think.

Why streaming loudness changes the comparison

There is a second reason not to chase a reference's loudness: most listeners won't hear it at full level anyway. Spotify's artist page on loudness normalisation says it adjusts tracks to -14 LUFS (ITU 1770) on its default Normal setting, with Loud at -11 and Quiet at -19 for Premium listeners. A reference mastered far louder than that is turned down on playback, and Spotify's page adds that full albums are normalised as a unit so softer songs keep their relative level. The web player and third-party devices such as TVs and speakers don't normalise, so some people do hear the raw file.

The practical upshot: in the mix stage you are matching balance and tone at equal loudness, not loudness itself. Final loudness is a mastering decision, and our guide to what each streaming platform actually documents about loudness covers how to set that target.

A quick checklist

• Static mix first, then load references.
• Reference bypasses your master-bus chain.
• Lossless files where possible.
• Match short-term LUFS on equivalent sections, then fine-tune by ear.
• Consistent monitoring level.
• Short A/B switches, with breaks.
• Compare one element at a time.
• Two or three references chosen for the same job.

What we think

Referencing is often taught as "make yours sound like theirs". We think that framing does artists a disservice. A reference is there to stop your ears drifting in an untreated room at 2 am, not to tell you what your song should be. The most useful thing level matching does is remove the loudness illusion, and once that is gone, a lot of the "my mix sounds worse" feeling goes with it. Often your mix was fine; it was just quieter.

So use references as a sanity check, not a template. If your vocal sits louder than every reference and you still like it, that may be your sound, not a mistake. Match the level, look honestly at the five comparisons above, fix what is genuinely off, and keep what is deliberately yours. The release you are measuring against was made by people who trusted their taste too.