Podcast Audio Quality: AI-based Post-processing vs. Hardware

I’ve been podcasting since 2015 and got really into audio when the plague started. Like…too much.


I’ve been obsessed with podcast audio quality for years, and have been through so…many…iterations of my setup. I started with a Yeti (still a great mic). Did the Electrovoice. Did the Shure SM7B. And now I’m using a Neumann U87ai.

You always want the initial recording to be as clean as possible.

But that’s not what matters. What matters is the quality of the chain, especially as you start fixing crap that’s in the original recording.

Here’s the basic rule. The most important rule. Don’t fuck with it.

You want to create as positive an environment as possible for your initial recording. Best mic. Best room. Best room treatment. Best mic position. The least amount of noise. Etc. That’s like 95% of the battle.

The other 85% of the battle is cleaning up the trash that will inevitably happen despite trying really hard on that. Such as de-essing (removing the sibilant, soul-piercing, s-sounds), removing mouth noises, background hisses and hums, etc.

The Rodecaster Pro 2

This is a huge debate in the audiophile world as well.

So that brings us to this post. The traditional wisdom is to do as much of this as possible using hardware. Why? Because hardware is better than software. That’s the thinking anyway. And it was mostly true for a very long time. But we’re seeing that change I think, and not just in audio.

So I got the RODECASTER PRO 2 when it came out. It’s super rad. Love the thing. And I naturally wanted to re-optimize my podcast chain using it. In other words, drop all my software plugins (I use the iZotope stuff, which is excellent).

So I did that, and here’s the result:

Unsupervised Learning — Security, Tech, and AI in 10 minutes…

Get a weekly breakdown of what's happening in security and tech—and why it matters.

Hardware optimization using the Rodecaster Pro 2

I like that a lot. It’s clean, loud, and solid. But it’s not SUPER clean. I can still hear artifacts in there. If you can’t hear them, listen to this one and you’ll see (👀👂?) it.

Software optimization using iZotope’s software plugins

If you listen to those on big speakers, in your car, and with headphones, you’ll hear that the second one is noticably cleaner.

And again—I’m doing very little to the natural sound of the Neumann U87ai. Just noise gate (for mouth and breathing noises), and a de-esser. And I did the extact same things on the RODECASTER.

I really wanted the hardware to win. I did. But it looks like software is just getting too good.

So if you’re in a similar position, consider doing this:

  1. Turn off EVERYTHING on your hardware. Just have the chain be as clean as possible

  2. Turn on the bass reduction on your mic, if it has that feature

  3. Use your favorite audio plugins to apply a noise gate for breathing and mouth noises

  4. Use your favorite audio plugins to take some aggro off your s’s

Don’t EQ it. Don’t add bass. You think it’s cool, but it produces fatigue in the listener and is hard to hear with low-end background noise like in a car or subway. And don’t worry about any other effects either.

Your voice is the instrument. Leave it as natural as possible. Let the mic do the work, and and let the hardware software do the rest.