Vocal Compression Settings for Rap (Starting Points That Work)

On a rap vocal, start at 4:1 with the threshold set for 2 to 4 dB of gain reduction on the loudest bars, attack 1 to 3 ms, release around 100 ms, and a sidechain high-pass at 80 Hz. Hand the rest to a second, slower compressor at 2:1 rather than pushing the first one harder. Per stage, 2 to 6 dB of gain reduction is the working range.

Disclosure: Ghostnote makes audio plugins. One of ours, TALISMAN, appears below as the worked example of a two-stage design and again in a labeled section at the end. Numbers credited to us come from its probe suite. No affiliate links.

What compressor settings should you start with on a rap vocal?

Every value below is a place to begin, and threshold is the only one you set by watching a meter. For numbers by flow, skip to the settings by delivery style.

How much compression does a rap vocal need?

Two to six dB per stage, and where you land depends on how even the take already is. Four steps:

1. Ride the clip gain first. Pull down the punch-ins and the one bar that sits above everything else until the waveform reads even. A take that needed 9 dB of gain reduction now needs 4.
2. Loop the loudest four bars and lower the threshold until the meter lands in the 2 to 6 dB range.
3. Watch the meter return to zero in the gaps. If it never does, the release is too slow and the compressor is holding the whole verse down.
4. Play the quietest verse last. If soft lines are squeezed as hard as loud ones, raise the threshold and let stage two do the leveling.

Why is one compressor not enough for rap vocals?

A rap take has two problems on two time scales. Syllable-level chaos: a hard consonant, a plosive on a "b", a punched-in line recorded an inch closer to the mic. Then phrase-level drift: the rapper leans in on bar seven, mutters the last line.

Set one compressor fast enough to catch the spikes and it chews the phrases into one loud texture. Set it gentle enough to level phrases and the spikes walk past it. One compressor has one attack time, which is why vocal chains have run two in series for decades: one box grabs transients in under a millisecond, the other releases slowly enough to hold a level across a bar.

Our own TALISMAN puts both stages in one plugin. GRIT is the FET side: instant peak detector, 4 dB knee, and a program-dependent ratio rise, so sustained input pushes the ratio up on its own (the all-buttons-in behavior). Its HEX mode reaches a 0.3 ms attack with ratios up to 16:1. WARMTH is the optical side: RMS detector, soft 9 dB knee, low ratios, two-stage program-dependent release. The gain-reduction meter reads full at 12 dB, well past where a rap vocal should sit.

Which compressor should you use for each stage?

Any two compressors work if their time constants differ. Engineers have paired a FET box with an optical one since the 1176 and the LA-2A, both now under the Universal Audio name. Stock plugins cover both roles: Logic's Compressor offers FET, VCA and opto circuit models, Ableton's switches between peak and RMS detection, and both manuals document the sidechain filter (Logic Pro User Guide, Ableton Live audio effect reference). For one clean digital compressor with knee, ratio and release all on the panel, FabFilter's Pro-C 2 is the usual pick and earns it. Order matters more than brand.

What attack time should you use on rap vocals?

Attack is a trade. Faster buys peak control and costs you the front edge of every consonant. That edge is diction, and rap lives on it.

Set an attack under a millisecond with several dB of gain reduction and you clamp every "t", "k" and "p". The vocal measures louder and sits farther back. A hook survives that; a dense sixteenth-note verse does not.

So keep the fast stage light. At 1 to 3 ms you catch the body of a shouted syllable while the tick of the consonant gets through. Leveling moves to the slow stage at 20 to 30 ms, where the gain arrives after the transient has left. Ad-libs are the exception: pin them at 0.3 to 1 ms and 8:1.

What release time should you use on rap vocals?

Release decides whether the compressor breathes with the flow or against it. Too fast and you hear it recover inside the gaps: room tone swells, breaths jump forward, the vocal pumps. Too slow and the line after the loud one arrives quiet.

Tie it to the tempo. At 140 BPM a sixteenth note runs about 107 ms, so 80 to 150 ms on the fast stage recovers roughly per word, and 200 to 300 ms on the slow stage recovers per phrase.

Auto release, sometimes labeled program-dependent release, does that arithmetic for you: short transients recover fast, sustained notes slowly, from one setting. Our WARMTH stage uses a two-stage version, and most stock compressors offer some form of it. For the topology underneath, look up the 2012 JAES compressor design tutorial by Giannoulis, Massberg and Reiss in the AES E-Library.

What ratio and knee work best on rap vocals?

Ratio and threshold are not independent, and treating them as separate dials is the mistake I correct most often. Ratio sets how hard the gain pulls once the signal is over the line, threshold sets where the line sits.

Two numbers from our probe suite make that concrete. On a −12 dBFS tone at a fixed threshold, 8:1 produced more than 4 dB of extra gain reduction compared with 2:1. Holding the ratio and raising the threshold from −30 dB to −6 dB removed more than 6 dB of it.

Working ranges for rap: 2:1 to 2.5:1 for leveling, 3:1 to 4:1 for peak catching, 8:1 and up when you want the compressor audible on ad-libs. Past 10:1 you are limiting.

Knee is the control nobody adjusts and everybody hears. A firm knee near 4 dB makes compression arrive abruptly at the threshold, so a peak stage feels like it is grabbing. A soft 9 dB knee starts working gradually below it, which is why optical leveling is hard to point at.

Why should you high-pass the compressor's sidechain?

The detector hears everything, and on a close rap vocal the loudest thing is often under 150 Hz: proximity effect at two inches, chest resonance, a hand on the stand. Leave it unfiltered and every plosive triggers full gain reduction, ducking the syllable behind it.

A sidechain high-pass filters what the detector listens to, not what comes out. Start at 80 Hz, or 100 to 120 Hz on a boomy close take. TALISMAN's sidechain high-pass sits at 80 Hz into both detectors by default: on a 60 Hz tone, opening that filter to 500 Hz removed more than 6 dB of detector-driven gain reduction. That is also what makes a vocal compressor usable on loops and 808s without the bass pumping the bus.

Is parallel compression worth it on rap vocals?

On melodic and half-sung deliveries, yes. On a dense verse, usually not: two stages of level control already did the job.

A parallel path raises what is quiet instead of holding down what is loud. Set the wet path heavy, 8:1 with a fast attack and 8 to 10 dB of gain reduction, then blend it under the dry at 20 to 40 percent. The peaks stay where the dry track left them.

What quietly ruins parallel setups is time alignment. Sum a signal with a delayed copy of itself and you get a comb filter, its first notch at half the sample rate divided by the delay in samples. Oversampling and lookahead on the wet path both add delay, and a few dozen samples of offset drops that notch into the low midrange where the vocal keeps its body. It reads thin and phasey.

Check it in ten seconds: bypass the compression on the wet path, flip that path's polarity, and the two should cancel to silence. In TALISMAN the parallel MIX path is latency-matched to the 4× oversampled saturation, so the blend cannot comb, and total latency is about 61 samples at 48 kHz.

How do you A/B a compressor honestly?

Level-match first, decide second. Louder wins every comparison it is allowed to enter, and compression plus makeup gain is a machine for making things louder. Streaming settles it: playback is loudness-normalized on delivery, as Spotify documents for artists. The measurement underneath is ITU-R BS.1770, the same algorithm behind broadcast's own loudness rules in EBU R128. Dynamics survive that normalization, a case Ian Shepherd has made for years. Gain does not.

So trim the output until bypass and engaged sit at the same level, then switch. Automatic level matching does it for you: TALISMAN's AUTO, on by default, does an RMS output-to-input match inside a ±12 dB window with 150 ms smoothing, so drive adds character instead of volume. Bobby Owsinski's mixing books cover the by-ear version of the same call.

Gain staging feeds all of it. Get the vocal peaking near −10 dBFS before the first compressor, because a threshold marked −18 dB, our default on both stages, only means what it says if the level arriving is in that neighborhood.

What compression settings work for each rap delivery style?

Delivery changes these numbers more than the microphone does. Where I start, after the clip-gain pass, with each stage cell reading ratio · attack · release · gain reduction:

Two neighboring decisions move these numbers. De-ess first, because a full-band detector reads sibilance as level and ducks the word behind an ess: the method is in how to remove sibilance without dulling the vocal. Tune after this stage rather than before it, with settings by delivery in autotune settings for rap. The full chain, stage by stage, is in the rap vocal chain in order.

From Our Own Line

TALISMAN is our vocal compressor and saturator, built around the two-stage approach above. Its ADVANCED panel exposes both stages, defaulting to GRIT at −18 dB threshold, ratio 4, attack 1.2 ms, release 110 ms, and WARMTH at −18 dB threshold, ratio 2.5, attack 25 ms, release 220 ms, with the sidechain high-pass at 80 Hz into both detectors.

After the compressor, saturation morphs from asymmetric-tanh tube through transformer to symmetric tape. TAME de-esses ahead of it at 5.5 kHz, −28 dB, 3.5:1, the second de-ess most rap vocals want anyway.

It is $29. VST3 on macOS and Windows, AU on macOS, keyless: no license server, no machine limit. No fake sales. No intro-price games. The price is the price. The rest of the line is on the Ghostnote plugins page.

Frequently Asked Questions

How much compression should I put on rap vocals?

Aim for 2 to 6 dB of gain reduction per stage, and split it across two compressors rather than asking one for 10 dB. A fast stage catches syllable peaks, a slow stage levels phrases. If the meter never returns to zero between phrases, you are compressing the performance out of the take.

What attack and release should I use on a rap vocal?

On the fast stage, 1 to 3 ms attack with 80 to 150 ms release. On the slow stage, 20 to 30 ms attack with 200 to 300 ms release, or auto. Faster attacks buy peak control and cost consonant detail, so keep the fast stage light and let the slow stage do the leveling.

Should I de-ess before or after compression?

Before. A compressor with a full-band detector hears sibilance as level and pulls the whole word down, so an unchecked ess makes the compressor duck the syllable behind it. De-ess, compress, then check the esses again after any saturation, because drive adds high-frequency energy.

Is 4:1 too high a ratio for vocals?

No. 4:1 is a normal peak-catching ratio when the threshold is set for only 2 to 4 dB of gain reduction. Ratio and threshold work together: our probe suite measured over 4 dB more gain reduction at 8:1 than at 2:1 on a −12 dBFS tone, so raise the threshold whenever you raise the ratio.

Should I tune the vocal before or after compressing it?

After. Cleanup, then compression, then saturation, then tuning, because a tuner tracks pitch more reliably on an even signal than on a take with 9 dB swings in it. The exception is hard-tune: when the effect is the sound of the record, print the tuned vocal first and compress that.