Q. How can I automate the level of vocals against a backing track?

Published in SOS December 2011
Sound Advice : Mixing

I’m currently involved in composing music tracks for a self-help medical audio program and I would like to be able to create tracks that combine subliminal messages with music. Subliminal messages work best when hardly heard — just a very little bit — so I’m looking for some automatic way of keeping the vocal signal just audible as the music rises and falls in level. I read your article about automatic ducking techniques in Cubase, and was hoping to use this information to achieve my goal, but I still haven’t been able to achieve what I’m after, namely to get the vocal level to follow the music level. I read the instructions several times, and played around with the settings, but it still doesn’t work. Can you advise?Via SOS web site

SOS contributor Mike Senior replies: The Cubase Notes article (SOS May 2009:  

www.soundonsound.com/sos/may09/articles/cubasetech_0509.htm) you’re referring to won’t directly help you to achieve what you want, because it gives instructions on how to reduce the level of one signal (the electric guitars in the article’s example) in response to another signal’s level increase (the lead vocal in the article’s example). A better alternative would be to follow the instructions in my June 2010 Cubase column (www.soundonsound.com/sos/jun10/articles/cubase_0610.htm), where I describe how to simulate the effects of Waves Vocal Rider using a side-chain-enabled Expander plug-in. 

In Mike Senior’s June 2010 Cubase column, we showed how to simulate the effects of Waves’ Vocal Rider using a side-chain-enabled Expander plug-in; we can see part of the process here. This technique automates the vocal to correspond with the level of the backing track: particularly useful for our reader, who is creating tracks that incorporate subliminal messages.

The down side of that approach, though, is that expanders with external side-chain access aren’t particularly common, so here’s an alternative scheme that uses a triggered compressor instead (these are more common):

Create a parallel channel fed from the vocal. (You could also just duplicate the vocal track, but this is a little less elegant because it makes later processing of the overall vocal tone less convenient.)

Compress the parallel channel.

Invert its polarity.

Now trigger its gain-reduction from the music channel. (If you have several music channels, then perhaps send them first to a group bus, so you can easily feed the compressor side-chain from there.)

Assuming that your plug-in delay compensation is working, this setup should give you the automatic level riding you’re after. When the mix is quiet the parallel channel is less compressed, and will cancel the lead vocal more (reducing its level), whereas, when the mix is loud the parallel channel will be more compressed and will cancel the lead vocal less (increasing its level).