Now a tricky part: Adding compression.
Compression makes loud sounds quieter. You pick a target volume, called the "threshold", and any sounds that are louder than the threshold are made quieter by a particular ratio. The greater the ratio, the more the sounds are reduced. Compressing at 2:1 will reduce things a bit, and compressing at 10:1 will even things out more aggressively.
After you've made the loud sounds quiet, you can increase the volume of everything (which is why it also seems to make quiet sounds louder).
Before we add compression, however, we want to include a meter to show how loud things are. (Fun fact: loudness is different from volume. My understanding is that loudness is perceptual, where as volume is mathematical.) Because podcasts are listened to in suboptimal listening conditions—in cars, in subways, through cheap speakers—it's generally recommended that podcasts aim for -18 LUFS. (LUFS are a kind of loudness unit..)
Here's how to add a loudness meter: Remember how I mentioned that the right-hand column is for the output track? By default, there's a hidden track which is your final Stereo Output—it's your last chance to make any changes or meter anything before the sound goes to your ears. So this is where we want to place our loudness meter.
Add an Audio FX > Metering > Loudness Meter > Stereo.
The Loudness panel should automatically pop up. (If it doesn't, just click on the Loudness Audio FX.)
The yellow line on the panel is a convenient target, and you can drag it to wherever you want. Set it to -18.
Notice that, because this is metering the Stereo Out, this will meter every track. Which is what you want.
And now the fun bit.
Add a compression Audio FX to each podcast guest’s track. (That'll be the left-hand column of audio effects!) Audio FX > Dynamics > Compressor > Stereo (or mono).
[The order. from top to bottom: De-noise, de-verb, EQ, compression. The effects are applied in order from top to bottom, and you want to get everything to sound right before compressing it.]
You'll want to locate two sections in the track you're working on—one where the speaker is saying something quietly (that you still want heard fully), and one where they're saying something loudly (laughs are good for this). Your goal is to fiddle with the knobs until both causing the loudness bars to dance around -18. It's totally fine if they go above it or below it a few notches, but you want them be focused on -18. Aim for speech to land between -12 and -24, but mostly lapping that -18 line.
There are three bars on the loudness meter, and I don't really know what they stand for, but let's say: M is for the current Moment, S is for Short-term, and I is for Integrated. That last bar only starts flickering once you've hit "Start" at the bottom, and it averages everything that has played since you hit "Start".
We're interested in keeping the middle one around -18 (or just under), and making sure the first one doesn't spike too high or too low.
Start by setting the main Threshold to -25, and the Ratio to 5. Set Auto Gain to -12 dB. Turn on the Limiter at the top and set its Threshold to, oh, -6 dB. And then listen, and tinker, until everything sounds about right.
Here, a video!
At the end of the video I discuss adding a compressor later in the process in order to make sure that, when several people are talking, they don't cumulatively make things too loud. For now, it's probably best to set up a compressor at a 1:1 ratio (i.e., no actual compression) with the Limiter on and a Threshold at... whatever seems reasonable. There is a fancier way to do this, though, which I'll get to eventually.
Have fun compressing!