Why your hand goes to the volume mid-podcast
The host is loud, the guest is quiet, and you turn it up and down the whole drive. Where that gap comes from, and why even good editing usually leaves it.
· 3 min read
In a lot of podcasts with two people, there's a moment when one hand leaves the wheel. The host talks loudly, the guest answers quietly, so you turn it up. Then the host blasts back through the speakers.
It happens so often that you stop noticing. Your hand goes to the knob on its own, the way it goes to the indicator.
Maybe it's the speaker?
The first thought is that something's wrong with the equipment: the car speaker, the Bluetooth, the phone. But play the same episode at home on good headphones and the gap is still there. It just bothers you less, because at home there's no road noise swallowing the quiet voice. In the car, any difference between the voices stands out more.
The gap is in the file itself, and it was there long before the episode reached your phone.
Every voice was recorded differently
The host sits in their own studio, close to a microphone they know, at a distance they've long since worked out. Even when the guest comes to the same studio, they sometimes sit further back or talk in a different direction. And sometimes they join over a remote call instead, on a laptop microphone, from a room with an echo.
Each of those two channels sounds fine when you listen to it alone. Put them together into one conversation and one of them is simply much louder than the other.
Why editing doesn't close it
In a lot of productions, editing takes care of the level of the episode as a whole. It makes sure the episode isn't too quiet compared with others, and that there are no sharp jumps in it. But it treats the episode as one unit.
The gap between two people taking turns to talk is a different problem. Closing it means handling each voice separately, through the whole episode, every time the speaker changes. Not every production gets to that step, and certainly not in every episode.
So what do you do
You can keep driving with your hand near the knob. Or you can hand that job to the player.
LevelCast listens to the episode as it plays and evens out the speakers as it goes, so the quiet voice comes up and the loud one is held back. The processing runs on the iPhone, and the audio comes straight from the producer's server without passing through anyone in the middle. There are three levelling strengths, off, normal and strong, and you can switch between them from the player.
To check that this works, we took a minute of "Two Microphones", a sample show made for LevelCast: Ido hosts from the studio, and Maya joins remotely and sounds quieter than him. The minute went through the app's levelling chain at the Strong setting. The gap between speakers dropped from 8.1 dB to 2.6 dB. The overall level actually went up. Before levelling it was -24.9 dBFS, and after it -13.7 dBFS. So the gap closed without the whole episode being turned down, and the two voices simply moved closer together.
You can hear both versions of that minute one after the other, with a switch that flips between them at the same second. Wait for the moment the guest starts talking, and switch right then.
If you're a producer
Levelling happens on the listener's side and doesn't touch your file. The download is still counted on your server, and the episode arrives exactly as your server sent it. All the details are on the page for producers.