Hello friends...and happy 2024!

I've been struggling with how to close out this past year.

If you've been paying any attention at all, you know that there's been a ton of doom and gloom in our industry as of late. And yeah, I could write about that. But seriously – who wants more doom and gloom?

And the truth is, it's been a very productive year for me. The project that consumed the bulk of my time – a whopping 64-percent of my 2023 work hours – was super rewarding. And fun fact: we were named one of The Atlantic's Best Podcasts of 2023! Go, team!

(Granted, we're #24 out of 25. But we're not last. We're second to last. And second to last is amazing.)

On the other hand, it's been a scary year to be a freelancer. My anchor client has disappeared. I have far fewer jobs lined up. I do have a really kick ass project that will keep me busy through mid-February. But after that? Who knows…

And job losses aren't the only angst that's been rocking our little audio world. There's also been a ton of drama over the coming of AI.

Will it take all of our jobs? Will it take over the world? Will we soon be bowing down to our AI overlords?

(Help me, Arnold Schwarzenegger. You're my only hope.)

But seriously, I use AI all the time. We used it extensively for Believable: The Coco Berthmann Story. And if you use transcription software, or run your audio through noise reduction filters, or...you know, Google things, then you are using AI.

So, how is a glass-half-full person supposed to close out this year? Hopefully with a message that conveys at least a little bit of hope?

The answer is to think of all the ways that I am – and will probably always be – better than AI.

Here we go.

#1: I Have Taste

Look, AI can generate a list of ideas. But it takes a human to know what of those ideas will appeal to other humans.

Longtime followers of this newsletter will undoubtedly know that I am a huge fan of Jessica Abel's Out on the Wire. There's a book and a podcast, but I process information audibly. So the podcast was the obvious choice for me!

Seriously, if you haven't checked it out, 2024 is your year.

I've never actually met Jessica, but I feel like we could be pals. Enough so that for a while, when I came across a storytelling dilemma, I would ask myself: WWJAD?

What Would Jessica Abel Do?

Anyway, Jessica has this rule: Follow Your Taste. And the idea is pretty simple. If you're inspired by something (intrigued by something, angered by something, disgusted by something) there's a pretty good chance that other people will be too.

So, go ahead. Use AI to generate a list of possible story ideas. But rest assured that the real talent comes in knowing what stories to pursue, how to frame them, and who to pitch them to.

AI cannot (yet) replicate that!

Curious about how to use AI to work more productively? Jeremy Caplan's Wonder Tools newsletter has lots of great ideas. His Best of 2023 list is a good place to start.

#2: I Have Empathy

There is soooo much empathy involved in storytelling.

And I'm not talking about the kind of empathy that makes you feel sorry for your subjects, ask them a bunch of softball questions and then use lots of melodramatic adjectives to describe the depths of their despair.

(There's really nothing I hate more than melodramatic adjectives. Blech!)

But I believe that my job as an interviewer is to understand my subject. Really and truly understand them.

I don't have to like them. I don't have to agree with them. But I really do need to understand.

That means that while I might feel badly for what they went through, I also challenge their perspectives. Because asking the hard questions is really the best way to get at the deeper truths.

I need to listen to the emotion behind my subject's words. Where do they hesitate? Where does their voice break?

These are the details worth following up on.

I need to know when to probe. When to keep quiet. When to offer a word of encouragement. And when to just sit in silence and let my subject come to me.

Wait for it...

I need to go past the "just the facts, ma'am" and find the soft underbelly of the story. It's not just about collecting information. I can get information from an email. Interviews are about collecting emotion, context and meaning.

All of this requires empathy.

AI can do incredible things, but it can't yet out-human a human. So let's lean into that!

#3: I Can Add Context

Information is at our fingertips…in ways it never has been before. But the ways that these details all connect? The meaning? The BIG picture?

Yeah, that (still) requires a human.

Look, I don't actually want a storyteller to tell me how to feel. I feel the way I feel, and if you try to police that, I get the icks.

But I do want a storyteller who intentionally chooses the information to include and the information to discard. Because breaking through the noise is hard work.

I want a storyteller to hold my hand and walk me through the complicated parts of the narrative, so I don't get lost.

I want a storyteller to help me find my way to "what it all means."

FYI -- that's very different from TELLING me what it all means. Show. Don't Tell.

Basically, I want a storyteller who can serve as a tour guide. Who can give me just enough context to help me find my way to the takeaway that's going to resonate in my life.

And, let's face it. Computers are not (quite) smart enough for that. Humans win again.

#4: I Can Hide Edits

2023 will be remembered as the year I really developed a love/hate relationship with Descript.

Descript is an incredibly powerful tool -- as are all the other AI powered, Descript-like programs that transcribe your audio and allow you to edit the text in order to edit the audio.

We used Descript a TON for Believable. I rough cut all our audio in Descript and assembled the selects. Another producer laid in the scratch tracks. And then, later, the studio tracks. We saved a ton of time and money working this way.

(And if we're being honest, we cost ourselves a ton of time and money. Because Descript can't yet handle a huge investigative project with hundreds of hours of audio. I mean, it's fine if you know from the outset how your episodes will be organized. But if you don't, you're kinda F-ed.)

But…back to my story. We assembled full episodes on Descript. But…those episodes are not what you'll hear if you listen to the podcast. We imported those assemblies into Pro Tools and fixed every single edit point.

AI cannot find a good edit point to save its life!

Podcasts that are edited by Descript – or a Descript-like program – hurt my ears. They sound sooo incredibly unnatural.

When I was first learning to edit, I was taught to edit to the beginnings of words. Always.

And it's not a bad place to start. If you always edit to the beginnings of the words, you will never clip off the end of the word. Or drop out in the middle of a breath! And the beginning of a word is generally easier to find than the end of a word.

But editing to the beginnings of words can be quite limiting. And the truth is, most edits require a little bit of finessing.

I love to hide my edits in the middle of an "s." S's can be soooo forgiving.

Also, a "t" is a "t." A "b" is a "b." A "p" is a "p." They're interchangeable.

If the end of a sentence sounds clipped, I'll go find another sentence that ends in the same word and use that instead. (Or the same syllable. Or sometimes just the same sound!)

If my reporter's voice cracks on the word "Afghanistan" in a podcast about Afghanistan, I'll go find another take and just replace the offending word.

I can hear when an edit sounds "right." And I'm gonna keep working it until I get there.

Speaking of which…

#5: I Can Hear Authenticity

Look, I loooove Adobe's new AI audio enhancer. It is sooo cool.

But, fun fact: AI utterly fails when given audio it can't comprehend. Like, for example when I entered a very compromised Zoom recording an Afghan women's rights activist.

She was in Afghaninstan and the interviewer was in Nairobi. Both of them had bad internet connections. So a ton of the interview was simply not usable.

But I decided to feed the entire interview into the program, hoping to tease out some portions of the interview that seemed like they might be okay, with a little enhancement.

And suddenly, in the middle of a woman interviewing another woman about women's rights in Afghanistan, the AI came across something it could not interpret. And that's how we ended up with audio of a male voice saying…

"Selfie!"

Yep. Selfie! Clear as day.

Why? How?

I do not know.

The truth is, the computer is going to try to make the audio as "perfect" as it can be. And in the process, it's going to create a lot of problems that never existed before.

So..two suggestions.

First, edit your audio BEFORE you put it through AI. Not only will that save a bunch of processing power, but it will also make sure that the computer is not trying to interpret useless data. And that should help avoid the worst of those crimes against audio.

And then, Adobe only gives you one control. It's a percentage – and the default is set at 90.

I want you to walk that back…all the way to 60% … or 50% … or even 40%.

Dial it back until you can hear the PERSON behind the AI.

This is one thing that AI really sucks at.

AI can strip out background noise.

It can add bass and treble and all the frequencies that are lost over Zoom.

It can make your audio sound PERFECT.

But it can't make your audio sound real. Because real is never perfect.

Real is messy. Real is awkward. Real is … real.

That's all for now! If you know me, you know that New Year's resolutions aren't really my thing. But I do hope to find time to write to y'all a bit more often this coming year.

And, as always, if you'd like to support Narrative Beat, please share this newsletter with friends. Or if you'd like to help offset my monthly Mailchimp bill while also having the chance to connect with other Narrative Beat readers, consider joining the Narrative Beat community.

Karen