Parallel compression blends a heavily compressed copy of a sound underneath the untouched original. You keep the natural attack and movement of the dry signal, and the crushed copy fills in the body and detail beneath it, so the part reads as fuller and denser without losing its life. It is also called New York compression.
The idea works the same way in almost any DAW and with almost any compressor: one dry signal, one hard compressed copy of it, mixed together. The amounts here are a starting zone, not fixed rules, and they depend on the source. None of this is tied to one product.
Parallel compression is a way of blending a heavily compressed copy of a signal underneath the original, untouched signal, instead of compressing the signal on its own. You end up hearing both at once: the dry signal keeps its natural transients and dynamics, and the squashed copy adds body, density and loudness underneath it.
Because the dry path is left alone, the loudest hits and the natural rise and fall of the sound come through exactly as they were. The parallel copy is compressed hard, which pulls its quiet detail up close to its loud parts, so it is dense and even. When you add that dense layer under the dry one, the quiet detail of the sound is lifted while the peaks stay where they were. The result is a part that is fuller and more solid but still breathes.
It is often called New York compression, after the studio engineers who made the trick well known on drums and vocals. The two names mean the same thing.
Parallel compression is not a special kind of compressor. It is a routing trick: any compressor will do, and the difference is that you mix its output back in under the dry sound rather than replacing the dry sound with it. Many modern compressors have a mix or blend knob that does exactly this inside one plugin.
Send the signal to a second, parallel path, compress that copy hard with a low threshold and a high ratio, then blend it in quietly underneath the dry signal until the part thickens. Start with the parallel copy silent and bring it up slowly, so you are always adding to the dry sound rather than swamping it.
There are two common ways to make the copy, and the second is usually tidier:
Once the copy is routed, compress it hard. These are the settings that matter and a sensible place to start, though the right amounts depend on the source and your taste.
| Control | What it does | A place to start |
|---|---|---|
| Threshold | Set it low so the compressor is working on most of the signal, not just the peaks. Parallel compression is meant to squeeze the copy, so the threshold usually sits well down. | Low enough for heavy gain reduction on the copy |
| Ratio | How hard the copy is turned down once it is over the threshold. A high ratio makes the copy dense and even, which is the point of the parallel layer. | 4:1 to 10:1 or higher |
| Attack | A faster attack catches the transients and makes the copy denser and more even; a slower attack lets some punch through the copy as well. | Medium to fast, by taste |
| Release | How quickly the copy recovers between hits. A faster release brings the body and sustain up, but too fast can pump. | Medium, or an auto setting, then adjust |
| Gain reduction | How hard the copy is being squashed. Unlike normal compression, heavy reduction here is expected, because it is only the copy being crushed. | Often 10 dB or more on the copy |
| Blend / send level | How much of the compressed copy sits under the dry signal. This is the main control, and the whole effect lives in how far you push it. | Start silent, raise until it thickens, then back off |
Blend it by ear. Bring the parallel level up from nothing until you hear the part get fuller and more solid, then ease it back a little so the dry signal still leads. If the compressed copy starts to dominate, you have gone too far and you lose the natural transients you were trying to keep. Compare against the dry alone at a matched level, so you are judging the effect itself and not just a jump in loudness.
Because you are mixing two versions of the same signal, anything that delays one copy relative to the other can cause phase cancellation, a thin or hollow sound as frequencies cancel. Some compressors, especially lookahead or linear-phase ones, add latency. On a send or aux path your DAW's plugin delay compensation usually lines things up for you; when you duplicate a track and compress the copy, check that the two stay aligned. Flip the blend to mono and listen: if it thins out or the low end weakens when you add the parallel copy, fix the timing before anything else.
Parallel compression shines anywhere you want more body and density without squashing the life out of a part: drums and the drum bus most of all, then vocals, and gently across a full mix. It is at its best on sources with strong transients and wide dynamics, because that is exactly what it lets you keep while still adding weight. The common jobs:
It does less where there are no transients to protect. A pad or a sustained synth that is already even gains little from it, and on a part that is already heavily compressed there is not much dynamic detail left to lift. Reach for it when a part has life you want to keep and body you want to add at the same time.
With normal compression the whole signal passes through the compressor and the compressed version replaces the original. With parallel compression the original stays untouched and you add a compressed copy underneath it. One replaces, the other adds, and that is the whole difference.
The consequence is what matters. Because normal, in-series compression is in the signal path, every peak and transient is turned down by the compressor, so the more you compress the more you flatten the natural dynamics. With parallel compression the dry path still carries the full, uncompressed transients, so you can compress the copy as hard as you like and still keep the snap of the original. You are lifting the quiet detail up toward the peaks, rather than pulling the peaks down toward the detail.
| Approach | What happens to the signal | Effect on dynamics |
|---|---|---|
| Normal compression | The whole signal passes through the compressor, so the compressed version replaces the original. | Turns the peaks down; pushed hard, the natural transients and dynamics flatten. |
| Parallel compression | The dry signal is left alone and a hard compressed copy is added underneath it. | Keeps the dry transients and lifts the quiet detail, so it adds body without flattening. |
If you are still getting comfortable with the controls on the compressor itself, threshold, ratio, attack and release, our guide to what a compressor does and how to use it covers them, and they behave the same way on the parallel copy. The only thing that changes with parallel compression is where the compressed sound goes: under the dry signal instead of in place of it.
A lot of compressors now have a mix, blend or dry/wet knob built in. Turning it below fully wet blends the uncompressed input back with the compressed output inside the one plugin, which is parallel compression without a separate send or duplicate track. It is the same idea done in one place, and a quick way to try the effect before you build a routing for it.
Parallel compression is a mixing move. It only matters once you have a part worth thickening, and writing and recording that part is the harder job. The routing on this page takes a few minutes to learn; having something worth pointing it at is the work.
The Collection is our three instruments together, ARGISH, SILT and REHEAT, with three separate licence keys. We do not make a compressor, and this page stays tool-agnostic on that. What we make is the source sound: ARGISH is a self-playing drone synth, SILT is a tape-loop instrument, and REHEAT writes acid lines. When the parts are clean and controlled to start with, a trick like parallel compression is a choice you make to add body, not a rescue for a sound that is fighting you. Every one ships as AU, VST3, AAX and a standalone app on macOS, signed and notarized, and a VST3 on Windows.
You can hear one free in your browser first, no install and no account, then decide.
We write these when there is something worth writing down. One email when a new one lands or a new Tunary instrument ships. No newsletter, no schedule.
Tunary