Audio editing order of operations (i.e. compressor, levels, normalize, etc)

As an audio engineer who has used Sound Forge, WaveLab (destructive editors), Pro Tools, Nuendo & Cubase, "Normalize" functions turn up or down the volume of a track equally to a set peak level. Normalize to -12db, and the track gets turned down so the highest peak in the file is -12db. It would not change the quiet sections to make them louder or vice versa.

EQ: Cutting all lows below 200hz can cut right into the human voice's range. Try 80hz and see if the rumble is gone, then adjust to taste. You don't want to steal a man's manly low end if he has a nice low voice.

Noise: Use noise reduction software before trying EQ. EQ is the hack way to do it, and it will take away from the voice. Or better yet, record in a quiet environment. At least get your mic away from your tower, turn off your heater and close your door. At the very least put the mic facing directly away from the loudest noise.

Varying voice levels: First in a NON DESTRUCTIVE EDITOR you can fix the major level issues by simply looking at the sound blobs. (You can even do that in Final Cut Pro by turning on waveforms.) See some that look a lot smaller than the rest? Make a cut on either side, and bring them up. But for the majority of level differences you can add some compression and smooth those levels out. There is also a plug called an "Ultramaximizer" or "Maximizer" which is a limiter with a volume boost combined. You set the limiter to the level you don't want to go above, then push up the volume. This will level out your voice, but it's more of a finishing plug you do last after you've done the other steps.

Recipes are cool, but they are like food recipes. Your own taste trumps any standard recipes. Use your ears. And EEK! Maybe don't fight with the people who answered when asked for help. There was nothing confrontational about the "EEKs". Just a quick way of saying those methods were off base for normal audio clean-up.
 
Tommy,

Well lets be fair here, your post came off a bit "I'm a rocket scientist, sit down and listen young man". Or at least that was how it seemed to me. So I wasn't feeling "classy" and I did think you were giving bad suggestions. I'm probably extra touchy on that since I all too often get handed mangled audio that someone has "cleaned up" and am expected to make it "good". And I thought I was doing it in a "friendly" manner with the EEK, but humor and the web...

So we don't get side tracked (LOL) I didn't mean to get harsh, and I kind of was, so apologies.

Now I will note that the formula you reprinted is significantly different than what you originally posted and is missing most of the things I objected to. It also talks about the normalizer properly, SF happens to have an optional limiter function tacked on (you didn't mention that). The formula as printed is actually quite reasonable. I would still probably not use the normalize but that is also probably habit and preference. And it seems to be aiming at a "pop music" standard of getting as close to zero as possible, and that is not generally what you want in film post (though maybe a good thing for podcast).

Breathing deeply...

Cheers
SK


Tommy K said:
Noize... wow ... that is all I can say to that dissertation above lol. I'm sorry you are having such a bad day. Sit down take a deep breath lol. I'm sure you are a great guy... You took some fairly harsh shots at me... not real classy on your part...
...

I am now posting word for word the recipe for low or on camera mike audio that was sent to me from a member of the Sound Forge development team. It is not my recipe... But it has been useful many times to me and others... If this applies to your particular situation, it works well for what it is intended...

...
Happy Forging!
Jim B
Click to expand...
 
budha said:
Hello everyone.

Is there a specific step by step process you should do when post processing narration/voice over audio as good as possible? I am recording podcasts for work now and I want them to sound great.

I understand you need a good mic, etc, but I am wondering what you do once it's recorded (I already have a good USB mic). I'm using STP3.

Also, is there a good way to make my voice sound richer/deeper, with an effect?

So for example (this probably isn't right, just throwing a list out there):

1. remove bg noise
2. adjust audio levels
3. run a limiter effect

Thoughts? Thanks!
Click to expand...

I think all the questions you're asking come from the "music" side of audio recording and less from dialogue recording for shows or motion picture.

If you want a track to sound good, record it well in a controlled environment, where recorded voice levels are higher than ambient noise around the talent.
All the processing tricks only come in handy where the location recording was done improperly or in a less that ideal environment.


After that something like an inflator, or a compressor set at ratio levels below 1.00 (on some plugins), better yet a multi band compressor like C4 or Multipressor in STP could help you push the "air" out of the monologue/dialogue - but it's just best to do it in a booth from the get-go.

All noise reduction algorithms are evil and should only be used when you cannot re-record or where the noise is so uniform that it actually helps and not hurts the signal - so a 60hz hum is easy to pick off, but the harddrive of your computer running while recording in a tinny echoey place with the microphone miles away from you - is not going to give you that radio announcer voice.

De-esser, EQ and other are your friends to make you voice over sound richer, but an MK416 MIC can do a lot more than a heavy studio condenser with a spit screen.

Audio Unit Plugins like Oxford Dynamics, Waves C4 help will help, so will Ren EQ and Ren Channel.

Don't use a normalize or a limiter or a finalizer of any kind.

And do not mix into a master bus compressor - Use a soft-knee compressor at a reasonable ratio on that track or bus itself only, after which use EQ to change tonal characteristics in the voice.

If you must (only if you must), use a GATE, tweak its settings and reaction time. - or instead of using a gate, use a stereo editor and just mute (x-fade out equal power 0db) the areas the talent is not speaking. Leaving breaths in is up to you.
 
Last edited:
I don't understand how a compressor can operated at ratios below 1.

My understanding that the ratio determines to relationship between gain in/gain out, and a ratio of 1:1 equals no compression, and the numbers go up from there (ie 2:1, 3:1, 4:1, etc). This would mean that as the signal approaches the threshold, the output gain would rise according to this ratio. In other words, a 2db rise in the source would result in a 1db rise in the output if set for 2:1. When I read a compression ratio of less than 1, I interpret that to mean expansion, which would emphasis the peakiness of a source meaning that a small gain increase in the source would result in a greater gain increase on the output. Kinda the opposite of what I'm looking for if I'm trying to "fatten" a voice.

Can you explain, please?

DJDecay said:
After that something like an inflator, or a compressor set at ratio levels below 1.00 (on some plugins), better yet a multi band compressor like C4 or Multipressor in STP could help you push the "air" out of the monologue/dialogue - but it's just best to do it in a booth from the get-go.
Click to expand...
 
I think technically what he's talking about is a Compressor/Expander. An Expander uses ratios less than 1 to expand the dynamic range and is often used instead of a gate to "expend" downward low level "noise". It's much smoother than gating.

Cheers
SK

gpforet said:
I don't understand how a compressor can operated at ratios below 1.

My understanding that the ratio determines to relationship between gain in/gain out, and a ratio of 1:1 equals no compression, and the numbers go up from there (ie 2:1, 3:1, 4:1, etc). This would mean that as the signal approaches the threshold, the output gain would rise according to this ratio. In other words, a 2db rise in the source would result in a 1db rise in the output if set for 2:1. When I read a compression ratio of less than 1, I interpret that to mean expansion, which would emphasis the peakiness of a source meaning that a small gain increase in the source would result in a greater gain increase on the output. Kinda the opposite of what I'm looking for if I'm trying to "fatten" a voice.

Can you explain, please?
Click to expand...
 
Yeah, what I meant is expander (late night and a early morning on deadline). I called it an "inflator" instead... The plug-ins I frequently use have an auto expander mode after 1.0:0 ratio. STP (the OP's daw of choice) has a dedicated expander plug-in. While Oxford dynamics, which I use frequently has a separate tab for it.

Being able to side-chain an EQ to an expander or compressor is also quite useful in many configurations as well.
 
Noiz2 said:
Tommy,

Well lets be fair here, your post came off a bit "I'm a rocket scientist, sit down and listen young man". Or at least that was how it seemed to me. So I wasn't feeling "classy" and I did think you were giving bad suggestions. I'm probably extra touchy on that since I all too often get handed mangled audio that someone has "cleaned up" and am expected to make it "good". And I thought I was doing it in a "friendly" manner with the EEK, but humor and the web...

So we don't get side tracked (LOL) I didn't mean to get harsh, and I kind of was, so apologies.

Now I will note that the formula you reprinted is significantly different than what you originally posted and is missing most of the things I objected to. It also talks about the normalizer properly, SF happens to have an optional limiter function tacked on (you didn't mention that). The formula as printed is actually quite reasonable. I would still probably not use the normalize but that is also probably habit and preference. And it seems to be aiming at a "pop music" standard of getting as close to zero as possible, and that is not generally what you want in film post (though maybe a good thing for podcast).

Breathing deeply...

Cheers
SK
Click to expand...

Thank you for your thoughtful post.. Apology certainly accepted. You are a good man:)

I offer my apology as well because I misinterpreted your intentions. I thought you were trying to embarrass me and my formula... this was not the case... that is clear now. My true intention is just to share, and try to give something back to this forum.. I certainly take a lot from it, and am better because of it... But I certainly wish to give only good advise ...

Mangled audio is definitely infuriating to say the least... I recently received video clips from someone where the audio was so low and so noise laden that it was impossible to clean it and use it... The women in the clips was speaking Russian to boot. I had to hire a Russian women to do ADR and replace the audio on the clips. I spent 3 days last week recording and trying to sync the ADR... It's now usable but not perfect... It was very hard to sync replace 15 minutes of interviews...


OK I now have a question that begs to be asked!

What is the proper scenario to use normalization functions? Peak and RMS.
It seems most of you are in agreement that it should be used only in certain instances or not at all. I would like to learn more about your opinions concerning normalization plugins.
 
That's how I understood it. But how do you "fatten" a voice with an expander? Typcially, when I want to fatten someone, I'm sqaushing them at about 4:1 with a threshold about 6db below where their waveform is peaking.

DJDecay said:
Yeah, what I meant is expander (late night and a early morning on deadline). I called it an "inflator" instead... The plug-ins I frequently use have an auto expander mode after 1.0:0 ratio. STP (the OP's daw of choice) has a dedicated expander plug-in. While Oxford dynamics, which I use frequently has a separate tab for it.

Being able to side-chain an EQ to an expander or compressor is also quite useful in many configurations as well.
Click to expand...
 
Tommy K said:
I spent 3 days last week recording and trying to sync the ADR... It's now usable but not perfect... It was very hard to sync replace 15 minutes of interviews...
Click to expand...

I have some ADR magic for you.

View the "Dialogue" demo and you'll be amazed: http://www.synchroarts.com/index.php?PAGEID=demos

And PS. Normalization is only a volume adjustment. The difference is it reads the top peak of the vile, and sets the top peak to the volume you wish, to prevent clipping. So for a quiet track you set normalization to, say, -3db. Then it turns up the file to the top peak is -3db. But be careful. If you have multiple clips from the same scene or session you want to normalize, they are not all going to have the same peaks, so normalizing them will result in audio that sounds like you changed the recording level between every scene.
 
Last edited:
gpforet said:
That's how I understood it. But how do you "fatten" a voice with an expander? Typcially, when I want to fatten someone, I'm sqaushing them at about 4:1 with a threshold about 6db below where their waveform is peaking.
Click to expand...

I use it to make the "room/air" between words on a dial track w/o PFX to sit lower on a db scale, it does create a "essshh..." effect at the end of a phrase, so a "room/air" track should also be there next to it on the submix to disguise it. 120 or sometimes 80ms Release. With a very tiny react time in ms. - Never for "fattening"

99% of my "fattening" is done with a Renaissance EQ/R-Channel or another non-transparent equalizer for narrative, for interviews I would use a transparent EQ, it does not need to sound dialog-e... Compression makes things sound too pumped. 2.5:1 is a safe bet as we're talking about speech and not a kick drum track sub mix.

The point is usually to give presence, making what the characters say intelligible, even if they are whispering or exhaling to show emotion - after that their location on the sound stage or illustrating positionally the room you see them in.

It also depends on the program material running next to it. So if you have a very bass-filled music queue, no matter how much bass you add to a voice from 53hz to say 160hz it's not going to matter during summing, it's going to compete with the music queue.

On something like an ENG interview track I'd only throw on a little glitz and that's if I cared enough.. Those need to sound real, as recorded, with only minimal processing to fit deliverable requirements for program levels - any additional processing aside from removing planes, traffic, car breaks, door slams, crew shoes and whispers won't service the material.
 
Great explanation, and yea, coming from a music mixing background old habits die hard. Thanks for giving more things to explore.

DJDecay said:
I use it to make the "room/air" between words on a dial track w/o PFX to sit lower on a db scale, it does create a "essshh..." effect at the end of a phrase, so a "room/air" track should also be there next to it on the submix to disguise it. 120 or sometimes 80ms Release. With a very tiny react time in ms. - Never for "fattening"

99% of my "fattening" is done with a Renaissance EQ/R-Channel or another non-transparent equalizer for narrative, for interviews I would use a transparent EQ, it does not need to sound dialog-e... Compression makes things sound too pumped. 2.5:1 is a safe bet as we're talking about speech and not a kick drum track sub mix.

The point is usually to give presence, making what the characters say intelligible, even if they are whispering or exhaling to show emotion - after that their location on the sound stage or illustrating positionally the room you see them in.

It also depends on the program material running next to it. So if you have a very bass-filled music queue, no matter how much bass you add to a voice from 53hz to say 160hz it's not going to matter during summing, it's going to compete with the music queue.

On something like an ENG interview track I'd only throw on a little glitz and that's if I cared enough.. Those need to sound real, as recorded, with only minimal processing to fit deliverable requirements for program levels - any additional processing aside from removing planes, traffic, car breaks, door slams, crew shoes and whispers won't service the material.
Click to expand...
 
It can be quirky but yes indeed Vocalign is a big tool in the tool chest. It also gets used a lot when your trying to steal lines from other takes.

Tommy, Normalization has a checkered past. My into to DAW's was basically theatre then film and we didn't need it for theatre and it was thought of as the devils seed in film so... But pop music has done it to death. The reason that it was though so bad of I think is that the CPU's were nothing by todays standards and the math gets complicated with normalizing and it had a tendency to trash the sound. Also it was done a lot with samples but when you do it to SFX and then layer up a bunch you have a problem. Also until fairly recently most/all "normalizers" set the peak to zero. Todays "normalizers" A) are using MUCH better algorithms B) are not the "just set the peak to zero" ones of the past. So probably the scorn tossed on them in film post is misplaced. It is essentially a over all gain that you don't have to set by trial and error.

The other part of the bad rep has to do with the way they are used. You will see posts that say something like "normalize to X and you don't have to mix". Dialog on the big screen wants a fair amount of dynamic range, AND it will be mixed with a lot of other sounds so you don't want it too hot. Mixers don't want a lot of dialog that is all "leveled" though they do want it balanced, which is close to the same thing. The closest thing I have seen to what people often use normalize and maximizers for but actually does it is "wave rider". I think it only works in Pro Tools though. It "looks" ahead and writes volume automation that tries to keep the track within a certain dynamic range. If you set it up correctly it does pretty much what a mixer would do.
It was really designed to do quick premixes on fast turnarounds like TV shows. I didn't end up getting it because for me it's quicker to just do it and it comes out better, I think. But it is an interesting idea that will probably become one of those "must have" plugins for TV and indi post places.

Wow, take a deep breath... ;~)

When would I use Normalize? I can't think of any time other than preparing samples for a sampler. But I do something very similar with a plugin in PT called "Gain". You can have it tell you what the peak or RMS level is of a region and then adjust the Gain and see where the output will end up. Functionally this should be the same as using a normalize set to that final value. One difference is you gan set "Gain" to peak at values above zero dB. WHAT, you say, there isn't anything above zero! True. THe above zero is just how much is going to be clipped. Sounds BAD, but actually on some things it really sounds good. If you have a hit that just doesn't sound hard enough. Clip it by 5 dB or so and you get some nice hardness. The key is you only clipping a VERY fast transient so it doesn't sound like a clip but it does sound hard.

When some clips are just TOO low to mix I have used gain on them and there is no reason other than habit to do that over normalizing.

So if I didn't have gain I would use normalize in those palaces. For "leveling" types of chores I use mixing. It REALY helps to have a control surface for that.
Cheers
SK

Chadfish said:
I have some ADR magic for you.

View the "Dialogue" demo and you'll be amazed: http://www.synchroarts.com/index.php?PAGEID=demos

And PS. Normalization is only a volume adjustment. The difference is it reads the top peak of the vile, and sets the top peak to the volume you wish, to prevent clipping. So for a quiet track you set normalization to, say, -3db. Then it turns up the file to the top peak is -3db. But be careful. If you have multiple clips from the same scene or session you want to normalize, they are not all going to have the same peaks, so normalizing them will result in audio that sounds like you changed the recording level between every scene.
Click to expand...
 
OK now you can say EEEK.

I more often get the blown out tracks than the low level stuff but a current short has some tracks so low I though they were just dead air. THey had NO waveform display. When you zoomed like crazy you got just a touch of variance on the flat line. They were using a shotgun at about 100 feet on a wide shot, no wireless.

But usually it's more like a clip I posted awhile back. Lowish level but a bad mic cable, a loose shock mount, the Refrigerator was going in the background and the boom op must have had palsy. It's a symphony of hums buzzes & bumps, with some dialog fighting to get through. And the actors were out of the country so no ADR.

So you have my sympathy.

Cheers
SK


Tommy K said:
...
Mangled audio is definitely infuriating to say the least... I recently received video clips from someone where the audio was so low and so noise laden that it was impossible to clean it and use it... The women in the clips was speaking Russian to boot. I had to hire a Russian women to do ADR and replace the audio on the clips. I spent 3 days last week recording and trying to sync the ADR... It's now usable but not perfect... It was very hard to sync replace 15 minutes of interviews...
Click to expand...
 
Hey Noiz2.

I have not yet purchased VocAlign, but plan to. You say it's quirky? How so? And is there something else that does the job better? I have my first ADR session coming up, and I want to get it right.
 
Noiz2 said:
Tommy, Normalization has a checkered past. My into to DAW's was basically theatre then film and we didn't need it for theatre and it was thought of as the devils seed in film so...
Click to expand...

How is 'normalize' done, vs. 'automatic gain control'... do the algorithms some how measure 'noise' then 'expand signals above that noise floor by some amount?
 
Chadfish said:
I have not yet purchased VocAlign, but plan to. You say it's quirky? How so? And is there something else that does the job better? I have my first ADR session coming up, and I want to get it right.
Click to expand...

Vocalign works well when the alternate production sound dialog or ADR line is close. If the line isn't close you may have to slice and dice a little before you Vocalign it, and yes it can be quirky, but I love it.
 
I'm going to get it as a VST plug for Cubase. I hear you have to make the 2 files the exact same length? Is that true in an NLE as a plug-in?
 
I think it's the best out there but it's not magic. It does a pretty good job but you need to go a little at a time. If you take a whole line it will tend to get wanky. This is kind of obvious when you think about what it's doing. But when you first get it the tendency is to treat it as a magic "fix" and then it can be disappointing. Like a lot of these things the less you need them the better they work.

It works and saves a LOT of time, but you still need to get pretty good ADR takes that are close, and it wont fix performance. In PT anyway the files don't need to be the same length. Actually if they were the same length you probably wouldn't need VocAlign...
 
You must log in or register to reply here.
Back
Top