Practical Mixing
How to use a reference track in REAPER (and level-match it first)
Published October 6, 2026
Import the released track onto its own track in your project, then measure both it and your mix in LUFS and pull the reference track’s fader down by the difference. REAPER can do the measuring with no plugins: there is an action that calculates the loudness of a selected item’s source media as a dry run, and the render dialog can report your mix’s loudness without writing a file. Until the two play at the same loudness, most of what you hear is a level difference rather than a mixing difference, and the louder one will seem better almost regardless of what is in it.
Comparing your mix to a record you love is one of the most useful things you can do, and one of the easiest to do wrong. Drag in a finished track, hit play, and your own mix sounds small, dull and far away. The usual conclusion is that something is missing from the mix. The usual reality is that the finished track is five to ten decibels louder, and at that point you are not comparing mixes at all.
The fix is short and it is arithmetic rather than taste. Measure both, lower the loud one, then listen. Everything after that section is about what to actually compare once the comparison is fair, and what a reference track cannot tell you no matter how carefully you match it.
Why an unmatched reference lies to you
Two separate things are working against you.
- Your ears do not have the same frequency response at every level. The normal equal-loudness-level contours, standardised in ISO 226:2023, map which combinations of frequency and sound pressure level listeners hear as equally loud. Those contours are not parallel: they flatten as level rises. In practice the same material played louder presents relatively more low and high frequency content, so a louder track sounds fuller at the bottom and more open at the top before anyone has mixed anything. The contours are measured with pure tones rather than music, so treat this as the direction of the effect, not a correction curve you can apply.
- The version you remember was probably loudness-normalised. Streaming services level playback out. Spotify states that it normalises tracks to -14 LUFS and advises keeping true peak below -1 dB, or below -2 dB for masters louder than -14 LUFS. So the track as you know it from an app has already been turned down to sit alongside everything else, while the file you just imported has not. Your memory of it and the file on your timeline are at two different levels.
Matching by dragging the reference fader until it “sounds about the same” is better than nothing but it is the thing the loudness difference is best at fooling. Spend the two minutes and get a number.
What you need
Stock REAPER, and a file you actually own. REAPER’s current version is 7.82, released October 4, 2026, and the measurement features used below arrived earlier in the 7.x line: version 7.80 added actions to calculate the stereo loudness of selected tracks via a dry run render, and 7.75 is the version whose changelog describes the equivalent action for the source media of selected items. If you are on an older build, the method is identical but you measure with a meter plugin instead. Our rundown of free LUFS meters for voice covers which ones are worth installing and when REAPER’s own measurement is already enough.
LUFS is the unit from ITU-R BS.1770, currently BS.1770-5, which specifies algorithms for measuring programme loudness and true-peak level. The integrated figure (LUFS-I) is one number describing a whole span of audio, which is exactly what you want for level matching.
The method, step by step
Pick a reference you know well, and own
Same style, similar instrumentation, similar vocal role. Use a file you bought or own, because you cannot import what a streaming app is playing. Two or three references beat one: a single track bakes in one engineer’s taste and one mastering decision.
Drop it on its own track at the top of the project
Its own track, not inside a folder or bus that is carrying processing. It should reach your monitors with nothing applied to it. Keep it muted by default so it never ends up in a render.
Take the master chain out of the comparison
If you have a limiter, EQ or anything else on the master, bypass the master FX chain while you A/B. Otherwise your master processing is being applied to the reference too, which quietly makes the reference sound like your mix.
Measure the reference
Select the reference item, open the Actions list, search for loudness, and run the action that calculates the loudness of the selected items’ source media via dry run render. It reports the figures without writing any media to disk. Write down the integrated value.
Measure your mix
Do a dry run render of the master with loudness calculation turned on and read the integrated figure. The controls are the same ones covered in our walkthrough of hitting a loudness target in REAPER’s render dialog. For a shorter comparison, set a time selection over one representative section and measure that instead.
Subtract, then lower the reference
Reference minus mix is your offset. If the reference reads -9 LUFS and your mix reads -17 LUFS, pull the reference track fader down by about 8 dB. Lower the loud one. Do not raise your mix to meet it, which only moves you toward clipping and changes what you are judging.
A/B in short bursts, one question at a time
Switch between the two every few seconds on the same section, and ask one question per pass. Comparing everything at once is how people end up making five changes and liking the result less.
Integrated loudness averages everything inside the span it measured. A full song against a thirty second loop of your chorus is not a like-for-like reading, because the song includes its quiet intro. Compare comparable sections: chorus to chorus, verse to verse.
What to compare, in order
Once the levels match, the differences that remain are real. Work from the largest to the smallest, because the big ones change your opinion of the small ones.
| What to check | How to hear it | What a difference usually means |
|---|---|---|
| Balance of lead voice against everything else | Switch between the two and track only how far forward the voice sits | Home mixes commonly land with the vocal too low, because you know the words already |
| Low end weight and where it stops | Listen for whether the bottom feels tight or vague, not for whether there is more of it | Often arrangement and a high-pass decision rather than EQ on the mix bus |
| Top end and air | Compare brightness on cymbals, consonants and breath | Could be mix EQ, could be the master. Treat this one with suspicion |
| Width | Listen to how far outside the speakers things sit | Usually a stereo content decision made during tracking and arrangement |
| Density and dynamics | Notice whether the reference stays constant while yours rises and falls | Normally compression and limiting, much of it applied in mastering |
Judging the lead voice is the one check worth doing every single time, and it has its own method. If you want a repeatable way to set that level rather than an impression, see how loud the vocal should be in the mix, which turns the balance into a number you can reproduce next session.
What a reference track will not do
This is where reference work goes wrong most often, so it is worth being blunt.
- It will not make your mix as loud as the record. Commercial loudness comes from mastering, mostly limiting, and it is a separate job done after the mix is finished. Chasing it with a limiter on the master while you are still mixing hides the problems you are trying to hear.
- It will not transfer the tone. A reference tells you a difference exists. It does not tell you which of a hundred decisions produced it, and copying a frequency curve from a finished master onto an unfinished mix usually makes things worse.
- It will not fix the arrangement. If the record has space around the voice because fewer things are playing, no processing on your side creates that space.
- It will not fix your room or your headphones. It does help you work around them, which is most of the value: a known track through an unknown system is still a fixed point.
- It cannot be reverse-engineered. The file you imported has been through a mix, a mastering chain and in many cases a format conversion. Some of what you are hearing cannot be produced from your session at all.
A level-matched reference will not make your mix sound like the record. It will stop you making changes you did not need, which is a bigger win than it sounds. Most of the time the honest result of a matched comparison is that the mix was closer than you thought and the vocal needed another decibel.
If the reference still sounds better
Work through this order before reaching for a plugin.
- Re-check the match. Measure again over the same section of both. An offset that was set by ear will usually be off by a few decibels, and a few decibels is the whole effect.
- Check the master is really bypassed. A limiter you forgot about flattens both tracks and makes the comparison meaningless.
- Separate mix from master. Density, glue and overall loudness are mastering traits. If every remaining difference is on that list, your mix may be fine and you are hearing the absence of a master.
- Try a second reference. If the difference disappears, you were comparing against one engineer’s preference rather than a standard.
- Check it is not the lead vocal itself. If the voice sounds thinner, duller or noisier than the record rather than simply quieter, that is a capture and chain question, not a balance question. Start with the stock REAPER vocal chain and come back to the reference afterwards.
A reference track is only worth anything after it is level-matched. Measure the reference with REAPER’s dry run loudness action, measure your mix with a dry run render, lower the reference by the difference, and then compare one thing at a time starting with how far forward the lead voice sits. Expect the result to be a shorter to-do list rather than a transformation.
If the vocal is what keeps losing the comparison
The most common honest outcome of a matched A/B is that the lead voice is quieter and less present than the record. The free StudioRescue vocal chain measures your vocal and sets REAPER’s stock plugins so you can hear a more finished version before you start making mix decisions. Stock plugins only, nothing to buy.
Get the free vocal chainSources
- Cockos, REAPER changelog - v7.80 (September 13, 2026) “Render: add actions to calculate stereo loudness of selected tracks via dry run render”; v7.75 (June 23, 2026) entries referencing “Calculate loudness of selected items source media via dry run render”. landoleet.org/whatsnew.txt
- Cockos, REAPER download page - current version 7.82, October 4, 2026. reaper.fm
- Cockos, REAPER User Guide - render dialog, loudness calculation and track routing. reaper.fm/userguide
- ITU-R BS.1770-5 (11/2023), “Algorithms to measure audio programme loudness and true-peak audio level”. itu.int
- ISO 226:2023 (3rd edition), “Acoustics - Normal equal-loudness-level contours”, covering 20 Hz to 12,500 Hz. iso.org
- Spotify for Artists, loudness normalization - normalises to -14 LUFS, recommends true peak below -1 dB, and below -2 dB for masters louder than -14 LUFS. support.spotify.com
All sources checked October 6, 2026. No listening tests are claimed; the loudness figures in the worked example are illustrative arithmetic, not measurements.