Still have a lot to learn but feels like we've turned a corner and are getting better sounding results for my video projects.
Compression of dialogue audio requires a bit of finesse to be effective. While learning, I've sometimes ended up with over compressed audio sounding something like this: Audio clip with a regular clip control law compressor
So first, why compress dialogue audio? So that the purpose of your film or video is served. The audio should help tell your story without making it hard for your audience. It allows you to loudness normalize to the established broadcast standards or recommendations so your audience doesn't strain to hear or experience the pain of audio that is too loud or distorted. For those that will watch or listen on mobile devices with earbuds, they can hear your project clearly.
We've covered audio compression in previous episodes but kept the description super simple as a start. That approach is a good intro but sometimes it doesn't work out very well and your audio ends up sounding over-compressed, sort of crushed and un-natural. So in this episode, let's look at a couple of things you can do to prevent that crushed sound.
First, check to see if you even need to compress your clip. Most of my footage needs just a little bit of compression so that I can loudness normalize for mobile and computer viewing. In audition, loudness normalize and see if your peaks get clipped. If not, no need to compress.
Compression graph: Axes = input level and output level and threshold
First and most obvious thing: Don't use a crazy compression ratio. There's no way it will sound natural. I like to stick to 2:1 per pass.
There's still a potential problem: Compression is applied in a pretty abrupt fashion, right at the threshold
One solution: Soft knee where the compression is applied more gently over a span of amplitudes (dBs, or loudness levels). And this can result in a more natural sound.
Unfortunately, Audition doesn't come with a compressor plugin that has a soft knee feature. Adobe, that'd be a really nice addition. :)
However, Audition supports VST plugins (or on Mac, both VST and AU). Here I'm using Izotope's Ozone 6 "Dynamics" plugin which is a really nice, though somewhat expensive option just to demonstrate how it works. There are gazillions of other options out there, some free, others at a price. If you have a favorite compressor plugin with a soft knee and an RMS "control law", let us know in the comments below! I've also included a link to one or two free options.
Speaking of "control law", what is a control law? This is how the compressor knows when to start compressing. Most compressors work based on peak control. They watch for peaks that hit or exceed the threshold and then start compressing. However, another option is an RMS control law. Without making the explanation too complex, this is essentially a way to average out the signal so that the compression starts its work in a more natural way--its isn't so fidgetty. This can also help to keep the dialogue sounding more natural vs. a peak based compressor.
Attack - how long the compressor waits after the waveform has exceed the threshold before it begins reducing the amplitude (or volume in laymen's terms). For dialogue, I keep this quite short, 20 ms or less, often right at 0 to 5 ms. We keep this short because we don't want too much of the peaks getting through since we are trying to loudness normalize.
Release - how quickly after the waveform has gone back under the threshold that the compressor stops reducing amplitude. I often set this between 80 and 150 ms. If you set too short a time, you'll bet a pumping sort of sound, and if set too long, you'll compress everything. Fancier compressors have an auto release which tracks the waveform and releases at the right time adaptively.
Then we loudness normalize and set to go!