← All posts · ·

Parallel Mastering: How to Blend Compression Without Losing Punch

Engineer balancing parallel compression in mastering room

Parallel mastering means sending your stereo mix down two paths, one untouched and one heavily processed, then blending them back together. The processed path is usually compression. You get the density and body of hard compression while the original peaks and transients stay largely intact. Done well, it lifts the quiet parts of a mix. Done carelessly, it adds phase problems, a lifted noise floor, and a track that only sounds better because it is louder.

This guide walks through the setup, the starting settings, the checks that catch the usual failures, and the situations where we would leave the technique alone.

TL;DR

Table of Contents

What parallel compression does in mastering

A normal compressor turns peaks down. Parallel compression works from the bottom up: you compress one copy very hard, then add it back beneath the unprocessed copy. The quiet details in the compressed path come up in level, while the peaks in the dry path still dominate the loud moments. Sound On Sound describes it as an arrangement that lifts quieter elements without crushing top-end dynamics or introducing the dulling effect that stacked compressors can produce.

On a stereo master, that translates into a few audible changes:

The trade-off is that you are adding the compressed signal on top of the original, so the overall level rises. That is why fair comparison matters, and we come back to it below.

How mastering differs from tracks and buses

On a drum bus or a single vocal, parallel compression is a sound-shaping tool. You can push it hard because you only affect one element, and the rest of the mix hides the artifacts. On the full stereo mix, everything is in the path at once. Every element you lift comes up together: cymbal wash, bass sustain, reverb, noise.

That changes how you work:

If you want the classic drum-bus version of this technique in detail, our guide to setting up and blending New York compression covers it.

Setting up the parallel path

There are two routes: a send to a second channel, or a dry/wet control inside a single processor. Sound On Sound notes that a separate compressed channel gives you room to add filtering, EQ, saturation, automation, and even several parallel chains. A built-in mix knob is quicker, but you give up that flexibility.

A basic send-based setup looks like this:

  1. Put your mastering chain's corrective processing first, so both paths receive the same tonal balance.
  2. Split the signal to a second bus, with the original path left clean.
  3. Insert a compressor on the second bus and set it aggressively (settings below).
  4. Bring the second bus up slowly under the dry path until the density changes, then stop.
  5. Check that both paths are in time with each other.
  6. Level-match your "on" and "off" states before judging.

The step-by-step routing details, including per-lane considerations, are in our parallel effects routing walkthrough.

Starting settings

Sound On Sound suggests a relatively modest 2:1 ratio, with the threshold adjusted so the loudest peaks see perhaps 20 dB of gain reduction. The same answer adds a practical condition: the compressor you choose should handle that much gain reduction without sounding nasty. Not every unit does.

Treat those numbers as a starting point and adjust from there:

Control Starting point What to listen for
Ratio 2:1 Gentle slope, so the compressed path does not choke
Threshold Low enough for roughly 20 dB of reduction on the loudest peaks Body and sustain coming up in quiet passages
Attack Moderate to slow Fast attacks can turn the sound mushy, so try slower first
Release Tied to the tempo of the material Pumping that breathes against the beat
Blend Start low and creep up Density without a change in the character of the transients

One note on attack: fast settings can make the compressed signal sound mushy, and parallel processing can reduce dynamic range while keeping more control over transients. That is one of the reasons the technique exists.

If you hear the mix pump, remember where the pumping comes from: the compressor's gain changes, set by threshold, ratio, attack and release. It is a dynamics-processing effect. If you hear it, change the compressor's timing or lower the amount of gain reduction rather than hunting for a fault elsewhere in the chain.

Shaping the compressed path

Because the compressed path is its own channel, you can treat it differently from the dry path. A common extension is a slight "smile" EQ on the compressed signal, described by Sound On Sound for drums, with a lift at both ends of the spectrum. On a master, keep that move small, and listen for whether you are adding air and weight or just brightness.

Saturation on the compressed path is another option. Be careful what you expect from it. Added harmonics change the perceived tone and density, but they do not give you free loudness: always check peaks and loudness meters rather than trusting your ears. Our 2–4 dB test for saturation before compression shows how to hear the difference in order of processing.

Phase and latency: the check that matters most

Parallel paths only sum cleanly when they arrive at the same time. If the compressed path is delayed relative to the dry path, the two combine with comb filtering, and the mix sounds hollow or phasey. Sound On Sound warns that phase problems arise in DAWs without accurate plug-in delay compensation, and that cancellation can rob transient-rich material of power.

A quick way to check:

  1. Flip the polarity of the compressed path and set it to the same level as the dry path.
  2. Listen or watch a meter. With perfectly aligned paths and a non-compressing setting, the sum should drop toward silence.
  3. If it does not, find the delay and compensate.

If the DAW's compensation does not fix it, Sound On Sound advises placing a short delay, often a handful of samples, in the direct path to match the processing delay of the compression path. Our walkthrough of sample-delay parallel processing goes through how to do that and how to confirm mono compatibility afterward.

Filters inside the parallel path

If you split the compressed path by frequency, the crossover itself can alter phase. Minimum-phase crossovers shift phase around the split point, while linear-phase options avoid that at the cost of added latency. Sound On Sound points to this exact trade-off in multiband processors: minimum-phase crossovers can cause phase shift, while linear-phase modes avoid unwanted coloration.

To be precise about the terms: a minimum-phase filter has the smallest delay for its magnitude response, but it is not delay-free. It still has group delay, and that varies with frequency. Linear-phase filters delay everything equally, so latency is higher. In either case, the two paths need to be aligned at the sum.

Crossover phase shift and latency comparison

If you split bands yourself in the DAW, a DAW-agnostic frequency split is a clean way to keep the bands phase-coherent. For parallel EQ specifically, see three DAW routes for parallel EQ.

Judging the result fairly

The most common mistake in parallel mastering is believing a louder version is a better version. Blended parallel processing raises average level, and a louder file wins most A/B tests by default. Mastering.com's mix-bus guidance makes the point that parallel processing can merely make a track louder, so compare against the original at equal level.

A fair comparison process:

On that last point, in mastering it can help to remove the parallel compression as a track fades, because it can slightly raise the noise floor as the audio dies away. Automating the blend toward zero at fade-outs is a simple fix.

When to skip it

Parallel mastering is not a repair tool. A useful reminder from mastering practitioners is to check whether the mix is already cohesive and balanced before you add compression, since mastering is polish, not repair. If the low end is unbalanced or a vocal is buried, a blend of compressed signal will only make the problem denser. A good habit is judging whether a mix needs corrective work by checking how the main elements sit against each other before you reach for any processing.

Other situations where we would hold back:

Going further: parallel by frequency band

A single full-range compressed path treats the whole spectrum the same way. Sometimes the problem is in one region: a thin midrange you want to thicken, or a low end you want to leave alone. A split into two or three bands lets you blend each region separately, so the lows stay tight while the mids gain density. Our guide to 2–3 band multiband parallel compression lays out per-band blends, and if time-based effects clutter your lows, frequency-specific delay uses the same logic.

Something else to keep in mind with stereo: Mid/Side compression can create apparent stereo expansion. Attenuating the mid image makes the sides sound louder relative to it. In a parallel context that means a mid-heavy compressed path can narrow the sides slightly, and a side-heavy one can widen. Check width after each change.

Perspective

We design plugins, and the problem we keep returning to is how clumsy parallel work is when you build it by hand. Bus sends, duplicated tracks, crossover plugins, delay compensation, level matching: every piece is a place for phase to go wrong, and every extra piece is another thing to check.

Our view is that parallel processing should behave like a single, coherent signal path. A frequency-specific approach, with each band on its own lane and its own blend, keeps the decisions where the ear makes them. You decide that the low mids need density and the top end does not, and the structure follows that decision rather than fighting it.

That is also why we care about real-time performance and low latency. Parallel paths only sum cleanly when processing is predictable and well aligned, and thoughtful DSP design is what makes that possible. The technique itself is old and simple. The craft is in keeping it phase-honest and judging it at matched level.

Kai

Try the technique with per-band lanes

If you want to build a per-band parallel setup without wiring buses and crossovers by hand, ToneLab is one way to do it. It puts multiple parallel effect lanes inside a single insert, with EQ targeting for each lane, and it is $73.99 as a one-time license with a free demo to try first. You can see details on the ToneLab pricing page.

Vector-dsp

FAQ

Is parallel compression the same as upward compression?

They sound similar in effect but work differently. Upward compression raises the level of quiet signals directly, while parallel compression gets there by blending a heavily compressed copy under the original. The parallel route keeps the dry peaks fully intact and gives you a single blend control to adjust.

How much of the compressed signal should I blend in?

Start with very little and increase it until you hear density and sustain change, then back off slightly. The right amount depends on the material, and on a full master it is usually less than on a drum bus. Match the levels of your on and off states before deciding, or you may mistake loudness for improvement.

Why does my master sound hollow after adding a parallel path?

The likely cause is a timing mismatch between the paths, producing comb filtering. Check that your DAW's delay compensation is active, then test with a polarity flip. If the paths still do not align, add a few samples of delay to the dry path to match the compressor's latency.

Should I use parallel compression on every master?

No. If the mix is already balanced, dense and dynamically controlled, there may be nothing for it to do. Reserve it for mixes that need body or low-level detail brought forward without flattening the peaks.

Sources