After completing the “comping and correction” stage, I now have five or so mono tracks of vocals ready for the next stage. I like to apply one last destructive edit before applying any FX or panning or mixing: “normalizing”. 

Audio Normalization is generally understood as making a collection of audio clips have the same peak value of 0 db. Many audio tools let you do this at the click of a button, but that’s not going to be helpful here. I need to do two or three things:

• ensure that the average level throughout each of the vocal tracks is constant, i.e it sounds like one consistent take;
• reduce breath sounds and noise between vocal phrases;
• make all vocal tracks sound the same volume when set at 0db gain.

I do this by adding a Gain Envelope to each clip, boosting and cutting where appropriate, and comparing across each of the five or so tracks:

on average, I’m generally boosting the tracks 3-6 db, and reducing the “intake breath before each line” by 6 db, and silencing anything else. When I solo these tracks, I can actually hear the bleed-through of the backing music from my headphones being picked up by the vocal microphone, so I make sure to replace those sections with silence.

After a final listen to each complete track to check for things I’ve missed, we get to the destructive part: For each track, I select all clips for an entire verse or chorus, and “bounce to clip”. This replaces the audio data with the new version, with the gain envelope applied:

That screenshot is of backing vocal tracks, showing the last phrase of a verse, followed by the chorus (hence the separate clips on each track).

I know it is unfair of me to talk about this process without providing audio samples, but I’m not quite secure enough for that. Even with pitch correction and gain normalization, these “naked” vocals are pretty unimpressive. Perhaps later.

Next: Mixing, routing, and effects.

A couple of days ago I thought I was done with “comping” the vocal takes, but on listening to the track all the way through, I became disatisfied with the melodic repetition in the choruses. A simple tweak to the bass in the second line made enough of a difference, working fine with the existing keyboard chords, but unfortunately the vocals clashed.

So I ended up re-recording the lead and backing vocals for that one line of each chorus. 

This morning I finished that task, so it’s on to the next stage.

With the keyboards complete, it’s time to move on to vocals. This is always a tricky process, and it is doubly frustrating this time, because I already have a perfectly servicable set of vocal tracks. It’s just that they are in the wrong key, and the lyrics are wrong. In other words, I have to do them all again.

For Abstracts, this is a total of 5 separate tracks: a single voice in the verses, doubled voice for the choruses, and a couple of voices of backing vocals throughout. Then at the end there are a couple more tracks on the final phrase, but we can ignore those for now.

Preparation for this involves setting up about 10 or 12 mono tracks with the input set to the microphone (Echo Layla Analog 1 for what it’s worth). Then, over a period of several days, I spend half an hour or so in the morning before work recording several takes of each verse/chorus, until I have at least 4 tracks of each lead and backing melody respectively. It takes me four or five days to get this completed. 

Why does it take so long? I’m not really a singer. I can sing in tune, kind of, but this is fairly high pitch for me and after a few runs through a verse, although I’m not hoarse, the timbre of my voice has changed to the point where it sound forced. So it’s time to stop for the day.

(If I had more training; pitched the song into a more comfortable range; and practiced more; I could probably speed this process up. But there’s not much chance of that.)

That brings us to the really tedious part: Selecting the “good” takes from each verse. For each take, I split the audio clip into sections, roughly equivalent to each line of the verse or chorus. The idea is that for each line, I now have four or five takes to choose from. 

This requires looping playback around each line, and selecting different takes to SOLO in order to identify the right performance to retain. The final vocal track can be made up of the best parts of each of the takes.

Some producers get quite obsessive about this, taking individual words from different takes. My vocal takes aren’t that bad! In the case of selecting a single vocal performance, there’s usually one take that stands out from the others. (When in doubt, choose Take 2.)

For doubled or backing vocals, it can get harder, because although there are tools to tweak the performance (more on that later), ideally you want the performances to match as closely as possible in terms of timing, phrasing, etc. Sometimes neither of the two that fit best together are the best take overall. So judgement is required.

Did I mention that I hate the sound of my singing voice? I’m sure I’m not alone in this. It can be a real block to this selection process, because you really need to be critical and listen carefully, and that’s hard to do when all you want to do is turn off the computer and go do something else. If you aren’t able to pay someone else to do this (and would you trust them if you could?) then you just have to get over it, and buckle down.

Let’s talk about AutoTune. The gimic of automatically locking the performance of a mediocre singer to absolute pitch is very distasteful to me. As David McLaughlin puts it, “Artists in the pre-AutoTune era HAD to be good. There wasn’t a record deal if you weren’t good. [..] Now I have a bunch of talentless clowns invading my ears.“

I don’t use AutoTune on my vocals. I do, however, apply pitch correction judiciously.

Pitch Correction is something that I can use to save time, turning the odd duff note in my vocal takes into something I can use, without having to go back and record another set of takes. If I didn’t have pitch correction tools available, I would have to record more takes, and spend more time cutting and pasting the takes to get an acceptable final result. I have actually done this in previous projects (see first paragraph) but since Cakewalk started bundling V-Vocal with SONAR 5.0, it’s been a hell of a lot less work.

The trick is just to not go crazy. It’s all about keeping the feel of the performance. On the other hand, some folks want to sound like robots. Whatever.

Despite the addition of pitch correction tools, it is still a long and tedious process. I’m currently halfway through Abstracts and it’ll be a few more sessions until I’m done.

Then we can talk about clip preparation, effects, and routing.

Apparently, Abstracts has always wanted a Hammond organ+GX1 layer on the keyboard side. Who could have guessed?

Three weeks ago I finished the drums for “Painting Abstracts” and moved on to the piano. I’m using Modartt’s Pianoteq Play instead of Truepianos this time around: for some reason it just seemed to work better. Both VSTs have a fine selection of “pianos” to choose from, but Pianoteq’s “K1” model seemed best. There’s a jazzy middle section coming up, however, that might prove the selection and see whether it sticks.

Last weekend I took a break from Abstracts and worked on “Strange Finale”, loading it up into Cakewalk’s SONAR X1 Producer and seeing how it behaved. Before long I was completely comfortable with the newly revised “smart” tool in the MIDI editor, and only once did X1 crash, taking out my recent edits.

Much has been written about the mad, crazy, and user-annoying changes between SONAR 8.5 and SONAR X1 so I won’t re-hash them. Suffice to say, I am not a fan of X1, and I’ve been cautious about using it full time for my projects. It’s a very buggy release, but in Cakewalk’s favour, they have been releasing regular patches to deal with the most troublesome niggles.

I just wish I could get SONAR X1’s “smart” editing tools into 8.5’s robust and stable UI.

*

In other news, lousy blog spammers still get through the captcha gate on this blogengine site, but it seems managable. I’ll keep comments open for a while longer.