Using a ducker to bring down the level of a music bed seems like the ideal solution to keeping a voiceover track clearly audible. We're all probably most familiar with this technique from its use in radio where the DJ's mic is used as the key signal feeding the side chain of a processor which brings down the level of the music passing through it. This example is probably responsible for the perception duckers sometimes have as a poor solution to this issue as heavy handed settings and over-enthusiastic DJs talking all over the music causing wild fluctuations in level are annoying.
This isn't as much an issue with the techniques as it is an example of inappropriate use and using a ducker is probably the best solution in a live environment but duckers have some issues which man that in other situations using automation, while more laborious is the only way to get acceptable results.
What Is The Problem With Duckers?
Assuming the ducker is set up appropriately there is one overriding issue with using them and thats that they are always late. Because they work "live" they can only attenuate the signal when the speaker has started speaking, it will always respond after the first word or so. This increases the temptation to use a fast fade out time which causes the ducking to sound more abrupt, and therefore noticeable. This can be alleviated to some extent by inserting a delay on the presenters mic so that the ducker can "look ahead" but this introduces latency which adds complexity and is only useable on a feed which isn't being monitored by the presenter.
Automation - Best Results, More Work
However tempting the automatic nature of ducking is, I never use it to dip VO in an edit. However carefully I set it up, it always needs some attention afterwards. Automation guarantees the correct result but it can take a long time.
This is where the stand-out feature of SpeakUp comes in. SpeakUp is a ducker, it works well in that role and offers simplicity meaning that a less technically minded presenter would have little trouble operating it but the real reason it genuinely extends the usefulness of a conventional ducker is that the gain reduction applied can be recorded as automation and edited after the automation pass. This reduces the burden of writing automation and means that just correcting the issues has to be dome manually, you're already 805 of the way there.
And of course the automation can be moved back to dip the level just before the presenter starts speaking using a standard nudge command.