Compressing vocals

octatonic

...


If I have a snare track that has a short snare without a lot of body to it I will usually want to elongate the snare.
I do this with a compressor.
I do not want to squash the entire signal- because if I use too short an attack then there will be no punch.
The rule of thumb is having too fast an attack will make drums sound further away, so will having too long a release.

I am not really using the compressor to level out the snare hits so they are all the same level.
The drummer does that by playing well (for the most part).
I am making a tonal change.

So I will set my threshold (or input) so that the loud snare hits trigger the compressor.
I set the ratio to something like 2:1 or 4:1.
I then adjust the attack so that the transient gets through.

Here are three files:

1. Snare with no compression on the channel (there is on the group and the master.)

https://jamesrichmond.com/audio/tfb/Snare No Compression.mp3

2. Snare with compression with moderate attack and release

https://jamesrichmond.com/audio/tfb/Snare moderate attack and release.mp3

3. Snare with compression with shortest attack and longest release

https://jamesrichmond.com/audio/tfb/Snare shortest attack longest release.mp3

Example 2 Is how I have the compressor (a 500 series Compex) set for the track I am working on.
Can you hear how it is noticeably thicker than the no compression example, with more tail. 
The snare is longer.
This is a good thing.
(I love the Compex as a snare compressor)

So in example 3 I've gone for the shortest attack and longest release.
Can you hear how it, frankly, sucks?

It has clamped down on everything.
Yes it is quieter- I didn't have time to level match, but even if they were it wouldn't be a good snare sound.
Having a moderate attack and release allows the snare level to come up overall, the tail to be extended but with still allowing the snare attack through.

'Transparent Compression':

A lot of the time I use compression as a tonal effect, rather than purely for levelling a signal.
For instance, you can adjust the release time of a compressor in parallel to make the ride pulse with the rhythm of the song.
Or you can use a Distressor or an 1176 on a room mic to crush it and then blend it in underneath the overheads to get a much bigger drum sound.

I do have transparent compressors.
The Crane Song STC8 can do 20dB of compression without really being heard and I have an Anamod Realios TL that can do it too.
But mostly I don't want a compressor to be completely transparent.

To your final point ' the slower the attack, the more likely we are to hear the compressor working - or is that a misconception??'

It's partially true, but it depends on the context.

How Attack Time Affects Compression Perception:
Slow Attack (e.g., 30ms - 100ms)

Allows transients (like drum hits, plucks, or consonants in vocals) to pass through before compression kicks in.
This can make the compression more obvious if it clamps down after the transient, causing a noticeable "pumping" or level drop. However, in some cases, a slow attack can sound more natural because it preserves the transient impact.

Fast Attack (e.g., 1ms - 10ms)

Immediately reduces the signal, which can smooth things out but may also dull transients if too aggressive.
This can make the compression less noticeable in some cases because it consistently controls dynamics.
However, extreme fast attack times can create a "squashed" sound, making the compression obvious in a different way.
So, is slower attack more noticeable?
Yes, if the compression clamps down too late, causing a noticeable level dip.
No, if used subtly, since it allows transients to shine while controlling dynamics naturally.
It really depends on the source material, threshold, ratio, and release settings.

Compression is my absolute favourite part of mixing. 
There is an art to it and i really took me over a decade of fucking around with different compressors on an almost daily basis to really get inside it.
Even now I am constantly surprised by what you can do with them, how they blend together.

As you can probably tell this is one of my favourite topics.
I'm a huge advocate for hardware compression.
I have all the plugin equivalents of the hardware compressors and I do use them but the hardware sounds different in almost every case.
(Notice I said 'different', not 'better'.)

Some get closer than others.
The UA 1176 plugins are mostly great.
I think every API 2500 emulation kinda sucks- it compresses but it does't sound much like my 2500.
The UA Distressors and Fatso emulations are good but I still prefer the hardware, especially at extreme settings.
Any Neve 2254 emulation doesn't sound anything like a 2254, or my BAE 10DCF's which are based on a 2254.

I am less picky about EQ's and for Reverbs I think plugins are pretty indistinguishable.

I've stayed up quite late writing this post, so I am sure I'll need to make corrections when I wake up, which I will do.
Hope this helps.

Stuckfast

Great stuff James. I am envious of your compressor collection! (Have you tried the McDSP APB compressors?)

One other point that might be worth making. Slowish attack and release times can often sound quite transparent because, in effect, the compressor never completes either the attack or the release phase. If you set both to, say, 500ms, then on paper, it should take half a second for the compressor to reach full gain reduction on a peak. But by the time it gets anywhere near that, the peak will have passed, and so it flips to the release phase before the 500ms is up. Then it encounters another peak before the release phase is complete. And so on. The upshot is that you get a fairly constant amount of gain reduction that varies quite slowly. 

octatonic

Stuckfast said:

Great stuff James. I am envious of your compressor collection! (Have you tried the McDSP APB compressors?)

One other point that might be worth making. Slowish attack and release times can often sound quite transparent because, in effect, the compressor never completes either the attack or the release phase. If you set both to, say, 500ms, then on paper, it should take half a second for the compressor to reach full gain reduction on a peak. But by the time it gets anywhere near that, the peak will have passed, and so it flips to the release phase before the 500ms is up. Then it encounters another peak before the release phase is complete. And so on. The upshot is that you get a fairly constant amount of gain reduction that varies quite slowly. 

Thanks Sam.

I haven't tried APB but I have all the McDSP plugins which are really great.
APB is on the list to demo at some point but I don't have £5k lying about right now.

Colin is an amazing dude but I am also slightly worried about ongoing support from small companies that have a single person (or a few individuals) doing most of the stuff.
Dave Hill died a few years ago and what happened to Crane Song?
A product that relies on a very small team for ongoing support for software updates is problematic for me.

It is why I don't have a Flock patcher too.
I love the idea but I just don't have confidence that 15 years from now that I will be able to use it with a modern OS, where my analogue outboard/patchbays will always work (and probably not crater in value).

octatonic

Here is another excellent video from Gregory Scott on hearing compression types.
This should really help a lot of people out.

digitalkettle

I miss UBK!

fastonebaz

This was all really interesting to read,  thanks for posting. 

Littlejonny

octatonic said:

...


If I have a snare track that has a short snare without a lot of body to it I will usually want to elongate the snare.
I do this with a compressor.
I do not want to squash the entire signal- because if I use too short an attack then there will be no punch.
The rule of thumb is having too fast an attack will make drums sound further away, so will having too long a release.

I am not really using the compressor to level out the snare hits so they are all the same level.
The drummer does that by playing well (for the most part).
I am making a tonal change.

So I will set my threshold (or input) so that the loud snare hits trigger the compressor.
I set the ratio to something like 2:1 or 4:1.
I then adjust the attack so that the transient gets through.

Here are three files:

1. Snare with no compression on the channel (there is on the group and the master.)

https://jamesrichmond.com/audio/tfb/Snare No Compression.mp3

2. Snare with compression with moderate attack and release

https://jamesrichmond.com/audio/tfb/Snare moderate attack and release.mp3

3. Snare with compression with shortest attack and longest release

https://jamesrichmond.com/audio/tfb/Snare shortest attack longest release.mp3

Example 2 Is how I have the compressor (a 500 series Compex) set for the track I am working on.
Can you hear how it is noticeably thicker than the no compression example, with more tail. 
The snare is longer.
This is a good thing.
(I love the Compex as a snare compressor)

So in example 3 I've gone for the shortest attack and longest release.
Can you hear how it, frankly, sucks?

It has clamped down on everything.
Yes it is quieter- I didn't have time to level match, but even if they were it wouldn't be a good snare sound.
Having a moderate attack and release allows the snare level to come up overall, the tail to be extended but with still allowing the snare attack through.

'Transparent Compression':

A lot of the time I use compression as a tonal effect, rather than purely for levelling a signal.
For instance, you can adjust the release time of a compressor in parallel to make the ride pulse with the rhythm of the song.
Or you can use a Distressor or an 1176 on a room mic to crush it and then blend it in underneath the overheads to get a much bigger drum sound.

I do have transparent compressors.
The Crane Song STC8 can do 20dB of compression without really being heard and I have an Anamod Realios TL that can do it too.
But mostly I don't want a compressor to be completely transparent.

To your final point ' the slower the attack, the more likely we are to hear the compressor working - or is that a misconception??'

It's partially true, but it depends on the context.

How Attack Time Affects Compression Perception:
Slow Attack (e.g., 30ms - 100ms)

Allows transients (like drum hits, plucks, or consonants in vocals) to pass through before compression kicks in.
This can make the compression more obvious if it clamps down after the transient, causing a noticeable "pumping" or level drop. However, in some cases, a slow attack can sound more natural because it preserves the transient impact.

Fast Attack (e.g., 1ms - 10ms)

Immediately reduces the signal, which can smooth things out but may also dull transients if too aggressive.
This can make the compression less noticeable in some cases because it consistently controls dynamics.
However, extreme fast attack times can create a "squashed" sound, making the compression obvious in a different way.
So, is slower attack more noticeable?
Yes, if the compression clamps down too late, causing a noticeable level dip.
No, if used subtly, since it allows transients to shine while controlling dynamics naturally.
It really depends on the source material, threshold, ratio, and release settings.

Compression is my absolute favourite part of mixing. 
There is an art to it and i really took me over a decade of fucking around with different compressors on an almost daily basis to really get inside it.
Even now I am constantly surprised by what you can do with them, how they blend together.

As you can probably tell this is one of my favourite topics.
I'm a huge advocate for hardware compression.
I have all the plugin equivalents of the hardware compressors and I do use them but the hardware sounds different in almost every case.
(Notice I said 'different', not 'better'.)

Some get closer than others.
The UA 1176 plugins are mostly great.
I think every API 2500 emulation kinda sucks- it compresses but it does't sound much like my 2500.
The UA Distressors and Fatso emulations are good but I still prefer the hardware, especially at extreme settings.
Any Neve 2254 emulation doesn't sound anything like a 2254, or my BAE 10DCF's which are based on a 2254.

I am less picky about EQ's and for Reverbs I think plugins are pretty indistinguishable.

I've stayed up quite late writing this post, so I am sure I'll need to make corrections when I wake up, which I will do.
Hope this helps.

I can tell this is a passion of yours! I’ve had a look at your studio website - bloody hell that’s some nice gear! 

I guess my original question is being slightly misinterpreted…I’ll phrase it differently: 

if you want transparent compression on vocals (as you might get with expert and meticulous fader riding) why is using a 0ms attack speed on your compresser not the way to go? If the attack speed is slower, do we not have a greater chance of hearing the compression?

i guess your wonderfully informative replies are really an impassioned explanation of why compression is great - I get that. 

When I compress my own voice, I like hearing the compression - it somehow makes me sound better. But with the singer in my band, she sounds better with very little audible compression. I don’t know exactly why this is, but I’m looking to level out the sound rather than give it punch.

mo6020

This thread is very interesting, thanks for all the great info chaps. 

octatonic

Littlejonny said:

I can tell this is a passion of yours! I’ve had a look at your studio website - bloody hell that’s some nice gear! 

I guess my original question is being slightly misinterpreted…I’ll phrase it differently: 

if you want transparent compression on vocals (as you might get with expert and meticulous fader riding) why is using a 0ms attack speed on your compresser not the way to go? If the attack speed is slower, do we not have a greater chance of hearing the compression?

A 0ms attack speed on a compressor can actually be counterproductive for transparent compression because it can clamp down on transients too aggressively, creating unnatural artefacts and potentially flattening the dynamics too much.

With a slightly slower attack time, you allow some of the natural transients of the vocal to come through before the compression kicks in. This can result in a more musical and natural-sounding compression, closer to how expert fader riding works, where the level adjustments are subtle and responsive rather than abrupt.

If the attack is too slow, you might start to hear the compression working, especially on peaks, but a well-chosen moderate attack time (e.g., 5-20ms) can maintain transparency while still controlling dynamics effectively.

Littlejonny said:

i guess your wonderfully informative replies are really an impassioned explanation of why compression is great - I get that. 

When I compress my own voice, I like hearing the compression - it somehow makes me sound better. But with the singer in my band, she sounds better with very little audible compression. I don’t know exactly why this is, but I’m looking to level out the sound rather than give it punch.

Compression can be great, when it is done well. 
Or it can ruin a performance, that is what I am getting at.

Different voices react differently to compression depending on factors like tone, dynamics, and even the way they interact with a mic. Some voices benefit from that ‘compressed’ sound, adding presence and character, while others sound more natural with minimal compression.

I'm not telling you not to use the fastest attack possible, but as someone who mixes audio, all day, every day, there are very few times I use the quickest attack time because my aesthetic is that the music sounds better when the transient isn't completely wrecked.

I am still levelling out the sound, I am just not making it dynamically flat.
Flat dynamics are boring to listen to.
FWIW, I didn't invent this stuff, this is just how records are made.

I would suggest doing this if you want a quick, simple approach without it being overcomplicated with multiple, serial compressors.

1. Set a medium-fast attack around 15ms and a medium release of around 80-100ms.
Don’t worry about the exact numbers.
Just think “medium”.

2. Set a moderate ratio of about 4:1 then bring down the threshold until you’re getting 10+dB of gain reduction on the volume peaks, enough so you can hear the compressor over-working.

3. Adjust the attack until you get the level of edge or forwardness for the genre. This is a taste thing. I prefer to let some of transient through. If you want to use the fastest attack every time, then no one is stopping you but if that makes your performance sound flat and un-dynamic then opening up the attack is how you fix that.

4. Adjust the release. Speed up the release if you want more aggressiveness or energy, slow it down if you want a more natural sound to the vocal.

5. Now reduce either the threshold and/or ratio to lower the amount of compression to about 4-5dB.

6. Make up the drop in level by adding some makeup gain.

Cirrus

Littlejonny said:

if you want transparent compression on vocals (as you might get with expert and meticulous fader riding) why is using a 0ms attack speed on your compresser not the way to go? If the attack speed is slower, do we not have a greater chance of hearing the compression?

A genuine 0ms attack time would make an immediate kink towards the top of every part of the wave that exceeds the threshold. There's another effect which does this that guitarists really like: distortion. Not always best for transparent compression!

Littlejonny

Thanks so much guys - there's so much good advice on this thread now...

On a similar note - does anyone here use Soothe or Babysmooth? It's a dynamic EQ and I was very impressed with the demo I had.

octatonic

Do you mean the product by Oeksound?

I have all the Oeksound plugins, I especially like Spiff.

Littlejonny

octatonic said:

Do you mean the product by Oeksound?

I have all the Oeksound plugins, I especially like Spiff.

Yes that’s the one

Stuckfast

A dynamic EQ like FabFilter Pro-MB is really useful on vocals. You can set it up so that it pushes the upper midrange on quiet passages but cuts it when the singer is belting, and it's great for controlling proximity effect.

Soothe is a bit different. It's a "dynamic resonance suppressor". More useful for controlling harshness. I have used it occasionally on vocals where someone has chosen the wrong mic for the job. 

I love Spiff. Bloom is amazing too.

octatonic

Stuckfast said:

A dynamic EQ like FabFilter Pro-MB is really useful on vocals. You can set it up so that it pushes the upper midrange on quiet passages but cuts it when the singer is belting, and it's great for controlling proximity effect.

Soothe is a bit different. It's a "dynamic resonance suppressor". More useful for controlling harshness. I have used it occasionally on vocals where someone has chosen the wrong mic for the job. 

I love Spiff. Bloom is amazing too.

I use Soothe a lot for taking the harshness out of guitars recorded with a modeller.
Another plugin which does a similar job, but differently is Gulfoss by Sound Theory.

I use dynamic EQ in Fab FIlter Pro Q4 a lot these days too.

PolarityMan

octatonic said:

3. Work the release.

Start by taking the release from longest to shortest- got back and forth a bit.
Watch how when you have a snare hit repeat, when you have the release long the compressor is still releasing the previous hot when it activates again.
Generally you do not want that.
When people set compressor release times they will usually want the compressor to have fully released before the next hit happens.
Now set the release to its longest and with each hit reduce it a bit and notice how the snare tail is lengthened.
This is important.

I would set the release around 100ms now.

Do you think there is any mileage in considering the tempo when setting the release? The idea just popped into my head. So for example at 160bpm 1/16ths are 94ms and if you know you have 1/16th repeated notes in the content this is going to bump into the release issue you mentioned.

Or do you think it's better to just optimise for the main hits and then do any automation required for snare rolls / fills / whatever?

PolarityMan

Littlejonny said:

My question I think is worth asking again - if you are clamping down AFTER the transient, then you aren't really controlling the dynamics, right? The transients are still varying in loudness even if what follows them is controlled?

If we want the compressor to flatten the dynamics of a signal, then why aren't we asking it to catch the transients? I'm aiming for transparent compression here - I don't want to hear the compressor working, rather act like an automatic volume adjustment. In my understanding, the slower the attack, the more likely we are to hear the compressor working - or is that a misconception??

For drums especially I've been getting more into using a combination of clipping and compression. Because the transient is so short on drums you can often clip it without any audible distortion which reduces the dynamic range and means you can get more body / less pokiness without compressing the transients. 

octatonic

PolarityMan said:

octatonic said:

3. Work the release.

Start by taking the release from longest to shortest- got back and forth a bit.
Watch how when you have a snare hit repeat, when you have the release long the compressor is still releasing the previous hot when it activates again.
Generally you do not want that.
When people set compressor release times they will usually want the compressor to have fully released before the next hit happens.
Now set the release to its longest and with each hit reduce it a bit and notice how the snare tail is lengthened.
This is important.

I would set the release around 100ms now.

Do you think there is any mileage in considering the tempo when setting the release? The idea just popped into my head. So for example at 160bpm 1/16ths are 94ms and if you know you have 1/16th repeated notes in the content this is going to bump into the release issue you mentioned.

Or do you think it's better to just optimise for the main hits and then do any automation required for snare rolls / fills / whatever?

Yes, all the time.
I set the release so that it is in time with the track.

This is basically what I meant when I said 'When people set compressor release times they will usually want the compressor to have fully released before the next hit happens.'

I don't usually have to do a calculation like that- just look at the meters and see if the compressor has released before it hits again.
There is also a 'vibe' element to it.
Sometimes the release time gives you a 'whoosh' type sound, especially on snare. If you get that in time with the track it makes the groove of the track more apparent.

The Gregory Scott video above talks about this with parallel compression.

Cirrus

PolarityMan said:

Littlejonny said:

My question I think is worth asking again - if you are clamping down AFTER the transient, then you aren't really controlling the dynamics, right? The transients are still varying in loudness even if what follows them is controlled?

If we want the compressor to flatten the dynamics of a signal, then why aren't we asking it to catch the transients? I'm aiming for transparent compression here - I don't want to hear the compressor working, rather act like an automatic volume adjustment. In my understanding, the slower the attack, the more likely we are to hear the compressor working - or is that a misconception??

For drums especially I've been getting more into using a combination of clipping and compression. Because the transient is so short on drums you can often clip it without any audible distortion which reduces the dynamic range and means you can get more body / less pokiness without compressing the transients. 

I'm forever looking for new ways to clip pleasingly. It's why I'm so fucked off with the Warm Audio Bus Comp I got last week - the release action is terrible, but somehow (possibly an accident of this particular unit) with the transformer output engaged it clips fantastically at about +20dBu which is *perfect* for bus saturation. So I don't want to sell it!

 I've been using 1081s for that job but they don't clip until like +30dbu which is way too hot for any following input stage without messing around with in line pads.

In the box there's a billion options, but I have been finding some combination of real amplifier and transformer saturation is more pleasing and immediate than shooting out 10 different plugins.

octatonic

Cirrus said:

PolarityMan said:

Littlejonny said:

My question I think is worth asking again - if you are clamping down AFTER the transient, then you aren't really controlling the dynamics, right? The transients are still varying in loudness even if what follows them is controlled?

If we want the compressor to flatten the dynamics of a signal, then why aren't we asking it to catch the transients? I'm aiming for transparent compression here - I don't want to hear the compressor working, rather act like an automatic volume adjustment. In my understanding, the slower the attack, the more likely we are to hear the compressor working - or is that a misconception??

For drums especially I've been getting more into using a combination of clipping and compression. Because the transient is so short on drums you can often clip it without any audible distortion which reduces the dynamic range and means you can get more body / less pokiness without compressing the transients. 

I'm forever looking for new ways to clip pleasingly. It's why I'm so fucked off with the Warm Audio Bus Comp I got last week - the release action is terrible, but somehow (possibly an accident of this particular unit) with the transformer output engaged it clips fantastically at about +20dBu which is *perfect* for bus saturation. So I don't want to sell it!

 I've been using 1081s for that job but they don't clip until like +30dbu which is way too hot for any following input stage without messing around with in line pads.

In the box there's a billion options, but I have been finding some combination of real amplifier and transformer saturation is more pleasing and immediate than shooting out 10 different plugins.

I needed some saturation for a guitar loop track earlier today.
I tired a bunch of different plugins and they were fine.
Each change I made was a different version of 'fine'.
Not blowing me away.

So I went to my Chandler Tube Driver which is on an insert, the rackmount one.
It sounds amazing in the context of the track.
Every time I made a change it was a different version of awesome.

That basically sums up digital vs analogue distortion for me.

I also hit a pair of Chandler Germanium preamps quite hard for a bit of saturation.