Sorry, I can't find the post from the other day - searching on here is broken - but I saw Craig Anderton mention he normalises phrases in vocal takes.  I've been pondering on this and can't understand why.  It can't be to get pseudo compression as it just raises parts of the audio by the same amount relative to the peak, and it also raises the noise floor in doing so.  I must be missing something here.

I edit vocal phrases by hand to get the volume consistently at the levels I want - so that compressors/EQ are working in a predictable fashion.  Many vocalists sing low notes more quietly, swallow opening/closing phrases, and take loud breaths - all of which I may want to adjust.   This editing might raise the noise floor, but even at 16 bit you're not going to notice this in the context of the mix. Some people choose to leave this levelling job to the compressor, (but you'll also get an increase in noise floor and breathing). Ultimately its about control I guess: if the vocal needs editing I want to be the one who is taking the decisions, and then use EQ/Compression etc - if it's actually required - on a top quality audio part.

I've done this in the past when recording my daughter's vocals - she's only just turned 6, so trying to get her to keep a constant distance from the mic is impossible.
 
Compressors are great for giving character to a voice, and I use them extensively, but for correcting massive volume differences in vocals just ends up crushing the loud bits way too much.
 
Splitting the vocals up into the various phrases and normalising each one gives a far more natural sound. I do this first, bounce to clips, then continue with my normal vocal processing (compression, eq, reverb etc).
 
Recently though, I've started using Waves Vocal Rider, which does the whole lot for me in a fraction of the time. It also saves me tearing out my hair due to the bugs in Cakewalk's clip handling. There are workarounds, but they're time consuming.
 

Been a while I recorded any vocals, but in the past I used volume automation to even things out then light compression.

jamesg1213
Been a while I recorded any vocals, but in the past I used volume automation to even things out then light compression.

I tried this, but volume automation affects the signal AFTER it's gone through compression etc. So massive differences in volume still hit the compressor at different levels, resulting in very different sounding results between phrases.
 
Making the signal more or less even before it hits any effects gets better results for me. It also means that I'm free to use volume automation later for real "mixing" duties rather than correction.
 

Both points are mentioned above.
1. If the vocals are consistent, they hit the compressor for the settings the compressor is geared for (another reason I normalize audio in general, then presets can be re-used).
2. A generic compressor works on a linear ratio, so evening out level will still possess a difference, just less depending on ratio set. Crushing audio to get that to nearly the same level can make the audio undesirable.

Bear in mind, compression on vocals should be rather light-handed.

was it this thread? Craig's first replies here. Google and the recent posts in a user profile are helpful search tools. Using US spelling helps depending on the user.

anderton
2. Phrase-by-phrase normalization can also add strength to the vocal.

msmcleod

jamesg1213
Been a while I recorded any vocals, but in the past I used volume automation to even things out then light compression.

I tried this, but volume automation affects the signal AFTER it's gone through compression etc. So massive differences in volume still hit the compressor at different levels, resulting in very different sounding results between phrases. 

 
You may be right in theory, but I never had that result.

msmcleod

jamesg1213
Been a while I recorded any vocals, but in the past I used volume automation to even things out then light compression.

I tried this, but volume automation affects the signal AFTER it's gone through compression etc. So massive differences in volume still hit the compressor at different levels, resulting in very different sounding results between phrases.
 
Making the signal more or less even before it hits any effects gets better results for me. It also means that I'm free to use volume automation later for real "mixing" duties rather than correction.
 

A good case for using Clip Gain Envelopes which get applied before any FX