A Little Advice On Input Compression Please?

Doctor Varney

Cave dwelling Luddite
Purpose: Narration for audio books. Digital recording.

Sorry if I've been here before but it seems a long time ago...

I record dry. I think this is generally accepted, good advice. Perhaps for obvious reasons of flexibility of mixing afterwards.

Intention
I wish to apply a little compression during input though, to help even out the levels. Even with my abilities in narration, I often produce a spike, word or sentence, possessed of more volume than the rest. This prevents the rest from normalising properly to 0dB before saving. I think some people would advise to add compression afterwards, in the mix. Except, I got rather confused by just aimlessly knob twiddling and listening to the results.

A Bit Of Background
My reasoning for ignoring that advice is that I have to manually attenuate those spikes in the waveform editor anyway - and then normalise. Unlike other media, such as music for my own pleasure, time is money and speed is of the essence for creativity's sake. So, I'm looking for a way of getting a fairly even waveform recorded to begin with. One that I can keep using, for all the many recordings I have to make.

After priming myself with info on how a comp works and what each control does, I've found a great way to help me get an idea of what the compressor is doing, in a visual way. Two audio editor plugins, either side of a compressor, allows me to record a dry, pre-recorded signal from one, into the other, after the compressor. It's helped me get the gist of what each specific control does, by comparing the shapes in the re-sampled waveform to the original. Sorry if that sounds a bit insane... but it was interesting! I like to experiment. :)

What I wish for, is a gentle evening out of the signal to create a lively performance, which is easy on the ears, which has enough volume throughout, without too many dips or spikes. What I don't want, is a perfectly flat waveform with no creative variation. That would just be awful!

So Onto My Main Question
I've learned enough to make sausages out of porcupines - and a few other things in between. Which is progress, but not ideal. I've got a result I quite like, by selecting a 'Vocal' preset on a multiband compressor plugin but, ideally, want to get better at controlling things manually.

Can anyone suggest settings that you'd normally use to achieve what I'm after?

Some common controls on various compressor plugins I have, include:

  • Threshold
  • Attack/ Release
  • Input gain
  • Ratio

I get what they are and what they do, but I'm not yet very good at using them all together. I can easily get lost in a knob twiddling frenzy and end up disappearing up my own arse.

Thank you very much!
 
Last edited:
The compression settings that work are going to be dependent on the source. Morgan Freeman's compression is going to be different than Sigourney Weaver's.

But for the moment, let's concentrate on those peaks --

G-Clip. Relatively shallow (90-92%), quite soft (20-25%) at whatever the gain needs to be at (which is going to be dependent on the input signal and the gain structure leading up to the plug).

Although in any case -- Almost universally, you're far better off doing 90% of your volume control and peak management "manually" (using volume automation, etc.). Compression doesn't just change the volume of the track -- It changes the dynamic range of the source (not just the apparent volume of the signal as a whole).

I'd stay away from maul-the-band compression unless there's a severe problem with the source signal (or you're using it to control sibilance or something, which should really be controlled at the source - and even then, something like SpitFish is going to be much more friendly).
 
Thank you, MM.

I can apply a volume envelope in the audio logger but I was rather hoping to get this done on the input to save a lot of time. Just mild compression. But I see what you mean about it squashing the dynamic range, rather than the volume. What about a limiter? Surely, there must be some sort of device which deals with just the volume?

Spitfish I use for it's intended purpose - as a de-esser. That's what I use to remove sibilance.

Massive Master said:
The compression settings that work are going to be dependent on the source. Morgan Freeman's compression is going to be different than Sigourney Weaver's.

I'm aware that settings would have to differ, depending on the artist's voice - but I'm interested to know what sort of settings to use on voices in general. And not just what (I can see by selecting a preset) but why.
 
Last edited:
At the input... As much as I wouldn't recommend it (especially with VO, where it's really going to be audible), I'd suggest something in the LA-2A-ish family. Manley/Langevin's DVC or even the ELOP would be VERY high on a very short list. Basic controls and "automatically nice" (for lack of a better term).

On the uber-budget-friendly, you could do worse than ART's Pro VLA if you can find a good unit (the good units are quite decent - But sometimes it takes several units to find "that" one) or even FMR's RNC.

Keep in mind that you have to pay extra-special attention to your signal level when compressing it at the input... Set your levels to "normal" BEFORE applying compression - and then don't even bother adding make-up gain (as it will almost undoubtedly be cleaner to add digitally). If you normally peak at around -12dBFS on the way in (which is a nice place to be), just let the compressor nip a few off and worry about it later.

All that aside -- Perhaps you might try something more designed for the purpose like a Symetrix unit...

Anyway -- There really aren't "usual" setting for voices. Everything matters. The speaker's dynamics and attack as much as the compressors threshold and release. An aggressive monotone staccato rant is going to have a totally different feel from a more "musical" and flowing presentation (think JFK vs. MLK).

Presets are worthless. And hardware isn't going to have any. And it's not like you can use a plugin at the input. So the best thing to do is to experiment with the tools at hand and find out what's right for your voice and your presentation.

If you want to throw something together quick, set your input level to give you a -10dBFS-ish peak on plosives and clicks which will probably put you at around -18dBFS average level (my personal favorite place to be and where your hardware is almost undoubtedly happiest). Put G-Clip on your last slot using the settings above and then add some sort of "optical-ish" compressor before it. Super long attack (100ms or more) fairly long release (3-4 seconds), low-ish ratio (1.5:1) and drag the threshold down until it's doing something interesting. Obviously, an amount of gain before the compressor is probably in order if it's going to be the only thing there and a certain amount after the compressor (using the comp's make-up-gain would be fine) before the G-Clip limiter --

But again, that's just a wild-guess but "broadcast-ish" starting point.
 
It seems like you intend to use a plugin compressor on the DAW channel and then record the resulting waveform, is that correct?

If so then there is no difference in recording dry and effecting after the recording is done except you don't get to hear the before and after.

I'd second the LA2A or LA3A type compressor as aside from being extremely good on vocal tracks they are also comparatively simple to use having only a compression and makeup gain control. the more you turn up the compression the more internally the compressor lowers the threshold and increases the ratio, additionally there are some great plugin versions of these compressors available from free to very expensive.

If you are talking really compressing on the input in hardware then LA2A or LA3A are expensive because they are fantastic and well regarded pieces. As Massive Master suggests, the ART Pro VLA can get LA2A ish in it' sound but it takes a lot more tweaking to get there as you have to set your thresholds, ratios, attack and release times by ear
the FMR RNC is also very transparent on vocals to in hardware but I never got along with it personally
 
It seems like you intend to use a plugin compressor on the DAW channel and then record the resulting waveform, is that correct?

Yes.

If so then there is no difference in recording dry and effecting after the recording is done except you don't get to hear the before and after.

True, but to save time when making dozens of short recordings in a day, I wish to have them down with the minimum of post-recording fuss and normalise and save them quickly. If I apply compression after the recording, I won't be able to normalise the waveform, before I save it.

I'd second the LA2A or LA3A type compressor as aside from being extremely good on vocal tracks they are also comparatively simple to use having only a compression and makeup gain control. the more you turn up the compression the more internally the compressor lowers the threshold and increases the ratio, additionally there are some great plugin versions of these compressors available from free to very expensive.

First, I'd like to know what 'LA2A etc' means. I'm relatively new to this and haven't a clue what you're talking about there. If I understand those terms, I think it'll be a lot easier for me. I have loads of compressor plugins - some are free and others, native, which came with my DAW software when I bought it. Many are purported to be extremely good but the problem is, as expressed in my original query, I don't really know how to use them. I'm hoping to get some tips and pointers on what settings get what results.

I have a native 'soft clipper' so I'm about to experiment with that.
 
And Doctor Varney -- PLEASE keep in mind that when using a plug-in on the input, any damage to the signal is permanent and usually hidden during recording. If you're overdriving your preamp, the plug isn't going to "fix it" on the way in - The damage is done long before the signal is even digitized, much less being routed through the plug.

If you want to set up a channel with all the bells & whistles & candy on it to monitor through, that's all fine - as long as everything about that signal is ideal and nominal at the input of that channel.
 
And Doctor Varney -- PLEASE keep in mind that when using a plug-in on the input, any damage to the signal is permanent and usually hidden during recording. If you're overdriving your preamp, the plug isn't going to "fix it" on the way in - The damage is done long before the signal is even digitized, much less being routed through the plug.

Wise words, MM and I would never disagree. Though, despite the problems I've been talking about in my gain structure recently, I believe I've come a long way towards sorting that problem. The signal is now strong and relatively noiseless, where before it was what you would call 'damaged'. It is now what I think I can expect of a dry signal, given my current gear situation and I have confidence that my target audience will approve. The desire to compress this otherwise satisfactory signal is born entirely out of the necessity to process my samples as quickly as possible, to meet the demands of my own deadlines. With all due respect to yourself (and the massive experience you have) I find that the general type of advice on forums like this, comes from a music making/ recording perspective. So I do my best, these days, to make sure the recording engineering gurus I consult, know up front, exactly what I'm trying to achieve and what level I'm currently playing at.
 
Here are the details on a real hardware LA2A compressor (peak reduction knob controls the overall compression)
Teletronix® LA-2A CLassic Leveling Amplifier | Universal Audio

Here's a good free plugin version (slightly more complex but there are others)
ThrillseekerLA – released today

Here's a good paid version plugin which only has the two controls
CLA-2A Compressor Limiter Plugin | Waves

Thanks, Bristol. That compressor sounds very cool, judging by it's blurb. May I ask - (and I don't mean to sound like I want to be spoon fed at the expense of your time) but, what settings would you use on that for male speech? And to get the effect I'm after (which is to even out the signal and prevent spikes)? Just generally, I mean.

I think, if I can get to a reasonable starting point, I'm sure I'll be able to modify the advice to deal with different types of voice.
 
Hey there,
If you get confused and end up in a knob twiddling frenzy, then a compressor on the way in is probably not a great idea.

If you can line up a VSt compressor and set in seconds, then one on the way in might be good, but if its a chore or a learning curve, you just stand to permanently ruin any recordings you make with hardware.

Sidestepping compression all together, why not look at something like Waves vocal rider.

It doesn't compress; It just turns down the volume when there are peaks (or up if there are troughs); It's automatic volume automation.
You set a threshold and the 'severity' of it, and boom, you're done.

I may have recommended this to you before, but if not, watch a youtube video tutorial of the digidesign stock compressor.
The graphical display makes it very easy to understand what all the knobs and settings do.
 
Thanks, Bristol. That compressor sounds very cool, judging by it's blurb. May I ask - (and I don't mean to sound like I want to be spoon fed at the expense of your time) but, what settings would you use on that for male speech? And to get the effect I'm after (which is to even out the signal and prevent spikes)? Just generally, I mean.

I think, if I can get to a reasonable starting point, I'm sure I'll be able to modify the advice to deal with different types of voice.

Hard to say since this is entirely level dependent

In the case of spoken word turn the peak reduction knob until the gain reduction VU is showing between 3 & 6 dB of reduction on the peaks only. try it in limiting mode too and see if higher compression but with a higher threshold sounds better. and then turn up the gain till the post compression level is where you want it

Tons of people make software emulations of this famous compressor and many have a free trial period (such as waves if you have an ilok) or native instruments called a VC2A if you don't.
NATIVE INSTRUMENTS
 
Hey there,
Sidestepping compression all together, why not look at something like Waves vocal rider.

It doesn't compress; It just turns down the volume when there are peaks (or up if there are troughs); It's automatic volume automation.
You set a threshold and the 'severity' of it, and boom, you're done.

Yes! It really IS just a case of pushing down peaks, in the volume sense. And, as MM said, a compressor reduces dynamic range. I'm glad he mentioned this.

So, I've looked about for a free volume rider, to no avail, and eavesdropped on forum conversations elsewhere. I've come to the conclusion that I can ride the wave manually, just like everyone else. I've tried it out and reckon I can get this sussed, no problem.

However... volume riding doesn't address the problem of recording a waveform that goes quiet in other places, due to a sudden spike. When normalised, those peaks are preventing the quieter parts from reaching my desired volume. So, I've decided to use a soft clipper (my DAWs answer to G-Clip) on the input.... JUST to tame those higher peaks. I figure a gentle threshold will preserve the subtle variations I want, in overall volume, as well as the dynamics.

Does that approach sound reasonable? The only reason I haven't recorded new samples in this manner yet is because I've discovered a noise removal tool which hopefully brings the work I've already done, up to scratch. My new gain settings will reduce the noise for future recordings. So this sound to you, like the right way to go?

Thanks for all your help so far, guys.
 
Last edited:
It's extremely unlikely compressing on the way in will give better results than doing it later unless you have spectacular gear and the ability to manage the compressor in real time during the performance.
 
However... volume riding doesn't address the problem of recording a waveform that goes quiet in other places, due to a sudden spike. When normalised, those peaks are preventing the quieter parts from reaching my desired volume.

It should. Just automate the loud bits to be quieter and the quiet bits to be louder.
Once you're in the ball park you should be able to just raise the volume to the desired level.
If you're still having a little trouble, then gentle compression or limiting might be the key.

If you get into manual fader riding, you can buy single fader controllers, like the presonus faderport.
 
However... volume riding doesn't address the problem of recording a waveform that goes quiet in other places, due to a sudden spike.

I don't understand. Just record so your peaks don't clip and knock them down later with a limiter or manually fix them, then gain up as needed. There are mastering limiters that will do this with one knob.
 
It's extremely unlikely compressing on the way in will give better results than doing it later unless you have spectacular gear and the ability to manage the compressor in real time during the performance.

Noted. Put to the test and discarded. I'm going with the advice here.

I don't understand. Just record so your peaks don't clip and knock them down later with a limiter or manually fix them, then gain up as needed. There are mastering limiters that will do this with one knob.

They didn't clip anyway. I can't actually get enough gain on the way in, for anything to clip! I was paranoid that the sound wouldn't be loud enough. If I manually lower a spike, then normalise, I get a more even waveform. Go too far with this and I'm making porcupine goulash.

I didn't want to manually attenuate spikes, simply because of the time constraint. As it turns out, I've actually spent more time messing with plugins, in the hope that I could save some presets for future quick use. :facepalm:

I tried a quick test project with volume riding automation. With -12dB headroom, turns out, in actual practice, the variation wasn't as stark as I thought and a gentle knock and push on the slider was all it took to even things out. Then, to get the mix up to 0dB, I put this pretty thing across the master bus...

p_t_MaximusScreenshot3_ezg_1.jpg

Then switched that 'mastering' tool off and plan to carry on tracking with my -14 to -12dB headroom... Until I'm ready to render the project.

It just means playing the whole project through with my finger on the slider. I can deal with that. It's quicker and cleaner than piling up plugins and twiddling knobs for hours and hours, only to possibly indelibly spoil my recordings on the way in.

If you get into manual fader riding, you can buy single fader controllers, like the presonus faderport.

I have this thing...
novation-zero-sl-mk-ii.jpg

I've spent so much time messing about, only to come back to the position of totally dry recording and moving a slider. I feel a bit stupid... but I've learned a valuable lesson.

Thank you for the help and support. I really appreciate it.

I have one more question about volume riding. Would you apply the automation to each chunk of recorded material or ride through the whole project?
 
Back
Top