Practical Mixing
How to A/B a plugin honestly in REAPER (level-match the bypass first)
Published November 13, 2026 · Updated November 15, 2026
If a plugin sounds better switched on than bypassed, first check that it is not simply louder, because the louder version of almost anything sounds better on a quick comparison. Loop one section, measure the loudness with the plugin on and with it bypassed, then turn the plugin’s own output control down (or up) until the two readings match. Only then flip the bypass and decide. In REAPER this takes a couple of minutes with stock tools, and it is the single most reliable way to stop adding processing that was never helping.
Most plugins change level as a side effect. An EQ boost adds energy, a compressor with make-up gain usually comes out louder than it went in, saturation adds harmonics and level, and even a de-esser or a cut can make the track quieter. When you click bypass, you are hearing two changes at once: the one the plugin was meant to make, and a volume change. Your ears are very good at noticing the second and very bad at separating it from the first.
The fix is not a better plugin or more experience. It is making the comparison fair before you trust it.
Why louder wins a bypass test
Two things push in the same direction.
- Hearing is not flat, and it changes with level. The normal equal-loudness-level contours, standardised in ISO 226:2023, show that how loud each frequency seems depends on how loud the sound is overall. Turn the same audio up and the low and high ends become relatively more audible, so the louder version tends to sound fuller and more open even though nothing about its tone has changed.
- Memory is short. Ethan Winer, writing on listening tests for ProSoundWeb, notes that hearing memory is poor and recommends switching quickly between the two versions. A long listen to one, then a long listen to the other, mostly compares your memory against the present.
Put those together and a plugin that adds a couple of decibels will usually “win” a casual A/B whether it improved the vocal or not. The opposite trap is just as real: a correct cut that makes the vocal a little quieter gets rejected because the bypassed version sounds bigger.
Do not match by peak level. A compressor lowers peaks while raising the average level, so two versions with the same peak reading can be a few decibels apart in loudness. Match the loudness reading, not the peak meter.
What you need
- The plugin you are testing, with its own output or gain control. ReaEQ has a Gain fader on the right of its window for output volume. In ReaComp the output gain control is labelled Wet. Most third-party plugins have an output, makeup or trim control.
- REAPER’s JS Analysis Loudness Meter (search
loudnessin the FX browser). It shows LUFS-M (momentary, a 400 ms window), LUFS-S (short-term, a 3 second window) and LUFS-I (integrated over everything it has measured). - A looped section of the song that represents the vocal: a full verse line or chorus, not a single word.
The method, step by step
Put the meter after the plugin
On the vocal track, add the loudness meter below the plugin you are testing, as the last item in the chain. Placed there, it reads the track with the plugin active and with it bypassed, so one meter measures both versions.
Loop a representative section
Set a time selection over 8 to 15 seconds of normal singing and turn on repeat. Use the same loop for every reading. A loud chorus and a quiet verse are different tests.
Read it bypassed, then active
Bypass the plugin with the checkbox next to its name in the FX chain (or the one in the top right corner of its window). Watch LUFS-S for two or three passes of the loop and note where it settles. Enable the plugin and do the same.
Trim the plugin’s output
If the active version reads higher, lower the plugin’s own output control by the difference. If it reads lower, raise it. Use the plugin’s control, not the track fader, because the fader would change both versions equally and fix nothing.
Re-check, then compare
Read both states again. As a practical starting point, aim for the two readings to be within about 0.5 dB of each other, closer if you can manage it. Then flip the bypass every few seconds on the same loop and ask one question: is the vocal better, or just different?
LUFS-S moves around as the vocal changes, so reading it by eye is approximate. For a precise figure, select the vocal item, open the Actions list, search for loudness and use the dry run action that calculates loudness including track FX. Run it once with the plugin bypassed and once active, and compare the integrated values. The same dry run approach is used to match a reference in how to set up and level-match a reference track in REAPER.
Hear only what the plugin is doing: delta solo
REAPER has a second tool that is useful here. Alt-click the small wet/dry knob next to a plugin’s UI button and REAPER switches to delta solo, which plays only the difference between the dry signal and the processed one. On an EQ you hear just the frequencies being added or removed. On a compressor you hear what is being taken off the peaks.
Use it after you have matched levels, not before. Any level difference the plugin adds also shows up in the difference signal, so on an unmatched plugin delta solo makes it sound as if far more is happening than really is. Once the levels match, it becomes a good reality check: if delta solo is nearly silent, the plugin is barely changing the sound and may not be worth keeping; if it is loud and full of the actual voice, you are processing more than you think. Alt-click the knob again to switch it off before you go on.
Testing a whole chain, not one plugin
Bypassing one plugin at a time is the right way to find a problem, as covered in why vocals sound robotic or over-processed in REAPER. But sometimes you want to know whether the whole chain beats the raw take. The level trim has to be part of what gets bypassed, otherwise the comparison is still unfair.
- In the FX chain window, select the plugins in the chain and choose Move to container from the FX menu. They now sit inside one FX container.
- Inside the container, add a ReaEQ as the last plugin, leave all its bands flat at 0 dB, and use only its Gain fader as a level trim.
- Put the loudness meter outside and after the container, then match the container on and off exactly as above.
The official REAPER guide describes FX containers as an advanced feature. If that is more than you need today, the simpler version is to bypass and level-match one plugin at a time, starting with the last one in the chain.
Make it blind if the decision matters
A level-matched test where you can see which button you clicked is much better than an unmatched one, but you still know which version you expect to prefer. For a decision you care about, take yourself out of it.
- Ask someone to toggle the bypass for you, in an order they choose, while you look away and say which version you prefer each time.
- Winer suggests that a listener should guess at least 10 to 20 times when trying to tell two versions apart. If your picks land close to half and half, the difference is too small to matter on this material.
- Switch quickly. A few seconds of each, on the same phrase, is easier to judge than long stretches.
| What you notice | Likely cause | Test | What to do |
|---|---|---|---|
| Every plugin you add sounds like an improvement | Each one is adding level | Loudness meter after the plugin, on and off | Trim each plugin’s output to match |
| A cut you made sounds worse | The cut made the track quieter | Match levels, then compare again | Raise the plugin’s output by the loss |
| Compressor sounds great, mix sounds squashed later | Make-up gain won the A/B | Peak and loudness readings disagree | Match loudness, not peaks |
| Matched, and delta solo is nearly silent | The plugin is barely changing the sound | Alt-click the wet/dry knob after matching | Consider removing it |
| Matched, but you still cannot decide | The difference is small | Blind toggling, 10 or more tries | Pick the simpler chain |
What a level-matched bypass will not do
- It will not tell you the plugin settings are right. It tells you whether the change is an improvement on this loop. A vocal can be better than bypassed and still be overdone. For starting points on the most common culprit, see vocal compression settings in REAPER.
- It will not survive a change in monitoring level. A matched comparison at one volume is a different test at another, because your hearing’s frequency balance shifts with level. Keep the listening level fixed while you compare, using how loud you should monitor when mixing as a guide.
- It will not judge the vocal in the mix for you. A plugin that sounds better soloed can sound worse against the beat, and the reverse. Do the final comparison with everything playing, and set the balance itself using how loud the vocal should sit in the mix.
- It will not fix the recording. If every option sounds poor, the problem is probably the take, the mic position or the room, not the plugin you are testing.
- Matching is approximate. Loudness meters model hearing; they do not replace it. Two versions at the same LUFS figure can still feel slightly different in loudness if their tone is very different. Within half a decibel or so, that is usually not what decides the test.
The usual result of doing this properly is not a revelation. It is a shorter chain. Some plugins that seemed essential turn out to be doing very little once the level difference is gone, and the vocal often sounds cleaner with fewer of them.
If something is not working
- The readings will not settle. Your loop has too much variation in it. Pick a shorter, steadier section, or use the dry run loudness action for an exact integrated number.
- The meter reads the same with the plugin on and off. Check the meter is after the plugin in the chain, not before it.
- You cannot find an output control. Put a ReaEQ with every band flat at 0 dB right after the plugin, use its Gain fader as the trim, and always bypass both of them at the same time, or use the container method above so one checkbox does it.
- Bypassing causes a click or a jump in timing. Compare on a loop and ignore the moment of switching. Some plugins add latency or take a moment to settle when they are switched back on.
A bypass test only tells you something once the level difference is gone. Meter after the plugin, loop one section, trim the plugin’s own output until on and off read the same, then switch quickly and, if it matters, blind. Keep the plugins that still win and remove the ones that only won by being louder.
Testing a vocal chain?
The free StudioRescue vocal chain uses REAPER’s stock plugins, so you can put the meter after it and run exactly this test on your own voice before you decide to keep it.
Get the free vocal chainSources
- Cockos, Up and Running: A REAPER User Guide v7.85: FX wet/dry mix and FX bypass; bypass checkbox in the FX chain and FX window; delta solo via Alt-click on the wet/dry control; ReaEQ Gain fader; ReaComp output gain labelled Wet; section 6.18 the JS Analysis Loudness Meter (LUFS-M 400 ms, LUFS-S 3 seconds, LUFS-I); dry run loudness calculation including track FX; section 17.6 FX containers and Move to container. dlx.reaper.fm/userguide/html
- ISO 226:2023, Acoustics: Normal equal-loudness-level contours (catalogue listing). din.de
- Ethan Winer, “Valid Comparisons: Conducting Software & Hardware Listening Tests,” ProSoundWeb, September 2024: fast switching because hearing memory is poor, fine-resolution level matching, and blind tests with 10 to 20 guesses. prosoundweb.com
- Sound & Vision, “The Problem with A/B Comparisons”: the ear’s sensitivity to volume changes in comparisons. soundandvision.com
- HomeRecording.com forum, “Parallel compression A/B level matching” (February 2024): a home mixer trying not to be fooled by a level increase when comparing a processed bus. homerecording.com
Levels and tolerances above are practical starting points, not published standards. No link above is an affiliate link.