I originally got into this technique with narration, where the "message" (whatever that might be) is the absolute most important part of a production. (The Melodyne technique works well too, but I find the blobs a little more ambiguous than waveforms, and I have shortcuts and a workflow that work well with Sonar's track view.)
 
Bear in mind that what I do is fairly labor-intensive, but also effective. For example, I might normalize a phrase after the inhale so that the vocal is brought up, but not the inhale. This is a degree of precision that a tool like Vocal Rider can't do. Also, I don't always normalize to the same level, and sometimes I use gain changes instead (like bumping up the level 2 dB or so). The goal is to have a subjectively consistent level so that you can get a strong, present vocal with only light amounts of compression or limiting. 
 
I also normalize guitar tracks going into an amp sim so that the way a preset affects the guitar is more consistent - I don't have to re-adjust drive or other parameters for different presets. 

I agree with Chuck. Using clip envelopes for vocals seems like a no-brainer to me. It's not just about evening out the levels, it's also about improving the phrasing by emphasizing key syllables while de-emphasizing others and reducing sibilant sounds.

I get results that I like best with clip envelopes as well.  Subtle changes are best, with light compression following.  I've mixed without clip envelopes and it turned out fine, but with is better for me. 

TECHNIQUE
I've done this for years now and not that I consider myself a pro, I find the most important step is the actual recording process itself (microphone singing techniques). that must be accomplished first and foremost.
It begins with (In this order):
 
- The correct mic for the job
- Trained singer who knows how to sing into a mic
- Correct distance from capsule
- Pop filter/grill
- Environment (Room ambience)
 
GAIN AUTOMATION
Next is the gain, how hard your signal is hitting your FX.
I don't see this step as being as important as it used to. Almost every FX now a day has some sort of input control.
If your vocals are hitting the compressor too hard, turn down the input. This is why a compressor was designed.
Too many people now a day needle things to death (including myself) there is no need to automate 2 db. of gain.
There IS however a need to automate 2 db. of volume when trying to sit vocals in the mix. Gain does have its uses but not like it used to.
 
VOLUME AUTOMATION
The most critical of all. this is where it all happens. If your singer wasn't perfect (some parts are louder than others) this is where make up is done. It is also where some choose to control their sibilance issues (P's SS's K's).
Some choose to use a De-esser but I prefer to hand limit. De-essers can eat up frequencies  in the sparkle range. Though it is more work intensive, the finished project is much cleaner.
At first, one song will take an hour or more. after a great deal of experience, you begin to recognize wave forms as tiny little "zippers". These zippers can easily be notched out 3-6 db. depending on the severity.

chuckebaby
TECHNIQUE
I've done this for years now and not that I consider myself a pro, I find the most important step is the actual recording process itself (microphone singing techniques). that must be accomplished first and foremost.
It begins with (In this order):
 
- The correct mic for the job
- Trained singer who knows how to sing into a mic
- Correct distance from capsule
- Pop filter/grill
- Environment (Room ambience)
 
GAIN AUTOMATION
Next is the gain, how hard your signal is hitting your FX.
I don't see this step as being as important as it used to. Almost every FX now a day has some sort of input control.
If your vocals are hitting the compressor too hard, turn down the input. This is why a compressor was designed.
Too many people now a day needle things to death (including myself) there is no need to automate 2 db. of gain.
There IS however a need to automate 2 db. of volume when trying to sit vocals in the mix. Gain does have its uses but not like it used to.
 
VOLUME AUTOMATION
The most critical of all. this is where it all happens. If your singer wasn't perfect (some parts are louder than others) this is where make up is done. It is also where some choose to control their sibilance issues (P's SS's K's).
Some choose to use a De-esser but I prefer to hand limit. De-essers can eat up frequencies  in the sparkle range. Though it is more work intensive, the finished project is much cleaner.
At first, one song will take an hour or more. after a great deal of experience, you begin to recognize wave forms as tiny little "zippers". These zippers can easily be notched out 3-6 db. depending on the severity.

100% agree with all you've said here.
 
I don't think gain automation / normalisation are mutually exclusive however.
 
I see Gain Automation as part of the mixing process, as you say, setting the level of how hard you want to hit the effects.
 
Normalisation on the other hand, along with removing "silent" parts from clips, deleting unwanted takes etc, I'd do on the cleanup stage between recording and mixing. I see anything that comes under destructive editing as being part of the cleanup, i.e. getting your tracks to the point where you want to start mixing.
 
Personally, I don't like using clip or volume automation to remove unwanted parts of a session. Anything that I'm not going to use (like bum notes, breathing during silent parts etc), should just be fixed or deleted. Massive (and I do mean massive, rather than subtle) variations in volume, I fix on the track, bounce to clips, and commit to the change. Same goes for pitch correction, although I don't use this often.
 
Whilst you lose the ability to go back, I don't feel I've lost anything by committing myself at this stage. It also reduces the amount of automation/correctional effects needed at mix time.
 

Since around my first music PC in 2004 I have used Wave Lab to fix vocals, spoken voice and even full choirs.
 
As you work you simply highlight the part that needs attention.
You can "see" where the waveform is weak and were loud spikes are. You can use the analyzer tool to see where those sections are at to assist your eyes. You can use the normalize tool or the gain tool. When done the track will be a much higher average RMS level, nice and even and well below clipping.  It a lot of work but Wave Lab makes this a much easier task than any other software I've tried.

msmcleod
I don't think gain automation / normalisation are mutually exclusive however.
 
I see Gain Automation as part of the mixing process, as you say, setting the level of how hard you want to hit the effects.

 
It's also the only way to deal with a change over time. For example, if a word trails off, normalization or gain changes won't do the job; you need to use Clip (or level) automation to add an equal and opposite ramp up to the end. The main difference for me is envelopes are for extremely precise adjustments, normalization and gain changes are for broad strokes.
 

Whilst you lose the ability to go back, I don't feel I've lost anything by committing myself at this stage. It also reduces the amount of automation/correctional effects needed at mix time.

I agree. I figure if I listened to something and felt it needed fixing, then it needed fixing . If the fix doesn't turn out as intended, I undo and move on.

Cactus Music
It a lot of work but Wave Lab makes this a much easier task than any other software I've tried.

With Cakewalk, I have a set of keyboard shortcuts that really speed up the workflow. 

Anderton

msmcleod
 
Whilst you lose the ability to go back, I don't feel I've lost anything by committing myself at this stage. It also reduces the amount of automation/correctional effects needed at mix time.

I agree. I figure if I listened to something and felt it needed fixing, then it needed fixing . If the fix doesn't turn out as intended, I undo and move on.

Sorry boys. I never do any destructive editing  
Mr.Anderton, you are correct..you CAN always "undo the move". Well at least until you close the project...then what ?
I don't even like bouncing/rendering Melodyne clips.
 
But its another good example of how we each have our own way of doing things.
I can respect that. Some times im ignorant to learning other methods (set in my ways).
 
 

In my over 20 years of rendering audio files with Wave Lab I have never ever experienced any downgrade to the sound,, only the improvements made buy fixing what was not good to begin with. As always there is the original recording backed up that in the 1 in 1,000th chance something did get broken, I can always return to the original.