So What? The Future of AI Podcasting: Trends and Predictions - Trust Insights Marketing Analytics Consulting

So What? Marketing Analytics and Insights Live

airs every Thursday at 1 pm EST.

You can watch on YouTube Live. Be sure to subscribe and follow so you never miss an episode!

In this episode of So What? The Trust Insights weekly livestream, you’ll learn how AI podcasting will change the way you think about audio and video. You’ll discover tools that enhance your workflow, create AI voiceovers, and even generate AI music. Discover how AI podcasting tools can save you time and open up a world of creative possibilities. Learn the ethical considerations and potential pitfalls of AI podcasting as you explore these cutting-edge technologies.

Watch the video here:

So What? The Future of AI Podcasting: Trends and Predictions

Watch this video on YouTube

Can’t see anything? Watch it on YouTube here.

In this episode you’ll learn:

Mind-blowing emerging tech: AI-generated content, real-time listener adaptation, and the rise of synthetic voices.
The human-AI partnership: We’ll share our thoughts on how AI might reshape podcasting as a profession.
Ethics roundtable: We’re tackling the big ethical questions surrounding AI’s role in the podcasting landscape.

Transcript:

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.

Christopher Penn – 0:29
Welcome to so what, the marketing, analytics, and insights live show. Chris and John here. Katie this week is in the hills of Vermont, mining exotic granite. So she will be back on the show next week.

John Wall – 0:43
There’s plenty of it up there, so that’s good. It should be a productive trip.

Christopher Penn – 0:48
Yep, this week. So this is the third and final part in our AI podcast series. In week one, we talked about improving your workflow using generative AI, particularly things like transcripts and show notes and stuff. If you missed that episode for some reason, you can always go back to our YouTube channel. Go to trustinsights.ai YouTube to catch up on back episodes. Last week we covered how to market your podcast more effectively. We looked at things like building a podcast audience profile and then taking things like a big marketing strategy plan and having generative AI help figure out what things made the most sense for your show. This week we’re going to go off in a slightly different direction.

We’re going to sort of, I guess, gaze into the crystal ball, such as it is, about the ways that podcasting is likely to be impacted with other technologies in generative AI, especially those in the voice and video realm, and talk about some of the other things like intellectual property, et cetera. So John, when have you played with any of the generative technologies in voice and video?

John Wall – 1:55
Yeah, I have a bit. But I was just definitely looking forward to this episode here because I know you’ve done a ton of stuff with this. Like, I’ve seen a couple few things that I can comment on, but I know you’ve been digging around deep in this bucket, so I am excited to kind of see what kind of stuff you played with. And then the other thing is not only is there the cool stuff you can do, but the implications of all of it. What productive can happen, and there’s just a gigantic mountain of ethical and other considerations that go into that, too.

Christopher Penn – 2:22
There are. In fact, you’re going to see this clip fairly frequently. This is one of my favorite clips from the movie Jurassic Park. Yeah, yeah. But your scientists were so preoccupied with…

John Wall – 2:30
Whether or not they could, they didn’t stop to think if they should. Yeah, that’s a, I’ve seen that a ton of times already, so.

Christopher Penn – 2:39
Exactly. All right, so let’s start with, let’s start with voice cloning because I think that’s probably one of the more interesting technologies, and particularly for podcasters, it’s something to think about. I’m going to go ahead and share my screen here, and there are a gazillion and a half voice cloning services out there. This is one called Eleven Labs. Probably the most well-known one. Relatively affordable. I’m on the starter plan, which is like $5 a month. And what you do is you give it a voice print, anywhere from a minute to 2 minutes of audio, and then you start processing it.

So let’s talk about how you would go about doing this if you were a podcaster. You might go to a recent episode of your show. So this is a recent episode of Marketing over Coffee. And you would download your polished MP3. You go into your show and then say, okay, well what? Maybe let’s try and just get a small snippet of this. So I have about a minute and a half of you talking here, John. We’re going to go ahead and load that into the Eleven Labs. And to do that, all you do is go into the Voices tab, add a voice to instant voice cloning and upload the voice.

Now, they do have a little checkmark saying that you have the rights and consents to use this. Nobody ever actually checks that while you’ve had a bunch of things like deepfakes of prominent political figures telling people not to bother voting. So anyway, yeah, you load that up, and then you get voices. And from there you can hat, you give it a piece of text.

Christopher Penn – 4:23
So this is for the old timers in podcasting. Once upon a time, a long time ago, we used to have promos, cross-show promos, where people would just sort of do like a station ID, you would say like, “Yo, this is John Wall from the Marketing over Coffee Podcast, marketingovercoffee.com. You’re listening to the Almost Timely Podcast with Christopher Penn.” And it was a fun way in the early days of podcasting for people to cross-promote on like shows. I don’t know that a ton of folks do that anymore, do they?

John Wall – 4:53
There’s a couple places where it happens, but yeah, it’s not like it used to be a huge thing, but this is hilarious. Now I totally get where you’re going with this. You can just be like, “This is Morgan Freeman. You’re listening to Marketing over Coffee.”

Christopher Penn – 5:05
Well, legally, you can’t impersonate somebody else without their permission, particularly for commercial purposes. I think that’s a really important thing to say. Like, yes, you can. Technically, from a technical perspective, you can load a voice that is not yours in here. Legally, you shouldn’t do that because that is, there’s any number of laws that will get you sued if you get caught. However, since, since we’re both here and live, I figured it’d be no big deal. So let’s hear what this actually sounds like. I’m going to hit play on this.

John Wall – 5:42
This is John Wall from the Marketing over Coffee Podcast at marketingovercoffee.com. You’re listening to the Almost Timely Podcast with Christopher Penn. Stay subscribed. This is John Wall from the Marketing over Coffee Podcast at…

Christopher Penn – 5:54
What do you think?

John Wall – 5:55
Yeah, that’s, that’s real close. I mean, I can tell there’s a difference, but the average person would probably not miss that, especially because with my own voice there’s very, it’s like I could. That seems to be trained a little closer when I was still having some sinus trouble about 2 weeks ago. And like, the variation between that and where I am now is the same dip. So yeah, that could easily pass as me for pretty much anything.

Christopher Penn – 6:20
It was trained off the May 30 episode.

John Wall – 6:23
Okay. Yeah, yeah. Sort of like a month back. Yeah.

Christopher Penn – 6:27
Which, when you think about it, is pretty crazy that it did a. Bear in mind, I did this setup literally 10 minutes before the show. So this is not, hours in advance. I literally, as we were getting ready for the show, I went over this thing, took a little snippet out in Adobe Audition, and stuffed it in here.

Now, if you think about it, that’s, a decent enough promo, particularly if you work out. You have some controls here for stability in terms of like, how close to the original source you want it to sound, how similar you want it to the original speaker, and so on and so forth. You can tweak those knobs. And once you get to a point where you’re happy with it and you’ve mastered it for that voice, you can start production.

When you think about it, if you wanted to do promos for a dozen, two dozen, three dozen shows, and send them out, that might be a handy time saver because you don’t want to read them all out loud. After a while, it gets kind of boring.

John Wall – 7:21
Yeah, yeah, exactly. To automate and have dozens of them cranked out, that’s an amazing time saver.

Christopher Penn – 7:28
Yep. And you could even do advertising. So you could, you could take some text like this and just have it say, “Are you tired of podcasts filled with fluff and maybe only one or two good ideas per hour?” and so on and so forth. And again, anything that the voice has knowledge of, the system can generate.

So here, let’s listen to this one. This one doesn’t sound like you quite as much, I’m pretty sure.

John Wall – 7:49
Are you tired of podcasts filled with fluff and maybe only one or two good ideas per hour? Tune into Marketing over Coffee every Friday with me and Christopher Penn. In 30 minutes or less each week, we’ll give you the rundown on what really matters in marketing. Are you tired of podcasts filled with fluff…?

Christopher Penn – 8:03
So that is the slightly longer one.

John Wall – 8:08
Yeah, yeah, that’s a little bit weirder. You can definitely, that’s uncanny valley territory we’re starting to wander into, but it’s still super. Like, especially if somebody was like, hey, send me 30 variants of what this could be like. Yeah, just cut and paste 30 times and throw them a file like that. That could be real.

Christopher Penn – 8:28
Exactly. Or as you’ve had happen many times, advertisers have sent you these long, convoluted scripts with all these tongue twisting words. You might want to say hi to do it.

John Wall – 8:42
Yeah, that’s interesting. And the other thing with that is, yeah, cause it could just give you some ideas on how to read. Because yeah, there’s always the challenge of when somebody gives you some gobbledygook, like, trying to make it sound like you would actually talk like that. So being able to have a couple different versions could definitely help.

Christopher Penn – 8:57
Exactly. Even if it was just, samples. You could just hand off the sample, say the advertisers, like, this is what’s going to sound like on the show, more or less.

John Wall – 9:04
Yeah. Here’s. And then the other thing is, yeah, I could give them alternative stuff, like, wouldn’t it better if it sounded like this?

Christopher Penn – 9:11
Exactly. So Eleven Labs, you pay per character. So you get 30,000 characters for $5 a month. And then obviously, every time you run the script, including repeats, it charges you characters as a deduction, your character balance, and ultimately, you can either buy more credits or upgrade to a faster plan.

That’s one thing I thought was really interesting about this, the technology. The technology is pretty decent for doing voice cloning. However, one of the new things that they’ve come up with, which I think is pretty cool, is the ability to sound effects.

So sound effects. Now, if you were to take some kind of common sound effects like a car screeching, let’s do a head, do a quick generation here. You could build yourself a little soundboard of effects that are derived from the, whatever latent space, the model has trained on.

So let me go ahead and download that. I have to load these things one at a time into StreamYard because it does not actually support playing audio. But let’s go ahead and see what does our car screeching sound like here? Where is it? It’s in my list. There we go.

John Wall – 10:29
It’s off in the distance, but it’s definitely there.

Christopher Penn – 10:32
Exactly. So if you wanted to build out a soundboard, particularly for a live show like this where you wanted to have some carts ready to go with, applause, or ringing bells, or whatever. Now with this tool, you can do that pretty easily.

John Wall – 10:48
Yeah, that’s the classic soundboard pile of stuff. And yeah, it is interesting the idea of, if you’re doing any kind of Foley stuff to do some kind of story to have an unlimited bucket of random sounds to go with. That’s great.

Christopher Penn – 11:01
Exactly. Now the other use case for this is especially if you want to do something in, say, a language you don’t speak. That, where now you start to get into really interesting territory. So I’m going to go ahead and go into Google’s Gemini here and load my standard translation prompt. And then I’m going to take this translation, the thing we were just doing, the copy for our promo, and I’m going to have Gemini do a good translation of that into Ukrainian. So we have our Ukrainian text here. Go back to Eleven Labs, put in our Ukrainian text, make sure it’s set on multilingual, we’re using your voice. Let’s make sure it increases similarity here, and let’s see how it generates. This will take about a minute or so. That was pretty fast.

And now let’s go and load this into StreamYard so that everyone can hear the John Wall promotion of Marketing over Coffee in Ukrainian.

John Wall – 12:19
(Speaks Ukrainian)

Christopher Penn – 12:31
So that is you in Ukrainian.

John Wall – 12:34
Yeah, that’s fantastic. It’d be interesting to play it for a native Ukrainian speaker and see how close it is to the mark.

Christopher Penn – 12:43
We, I mean, we know a bunch of those folks, so maybe we’ll tag them on LinkedIn and see what they have to say when the episode airs.

John Wall – 12:50
That sounds good. Yeah. And that’s fantastic too, because there’s so much stuff that goes into internationalization. And to be able to just take training materials or other stuff and get it translated without too much headache is fantastic.

Christopher Penn – 13:04
Exactly. And it gets, the thing I like about it is that it gets the pronunciation, but it is still 100% your voice.

John Wall – 13:11
Yeah. The same timber, same sound.

Christopher Penn – 13:15
Exactly. And they actually do have features in here for doing dubbing if you wanted to actually dub the target language. Their translations are not as good as Gemini, so I typically don’t use this. I will typically just use the Gemini text because one of the things that, one of the things really tricky about translation is there’s a lot of cultural context that doesn’t necessarily translate well.

So if you look at the results here in Gemini, it says explanation choices. Fluff was translated as into water because it’s a common Ukrainian idiom for unnecessary or meaningless content. That’s not going to happen in your standard translation software. You have to have that in your prompt to tell it. This is, these are the culturally appropriate things you want to translate.

John Wall – 14:01
Yeah. Because that’s where you see all this stuff go sideways. As far as, all these idioms we have in English, in American specifically, that if you translate them literally, it just sounds ridiculous.

Christopher Penn – 14:12
Exactly. So that’s Eleven Labs. That’s a lot of voice and sound effects stuff. If you wanted to add this to your toolkit as a podcaster, you have to use it judiciously. So one of the things we talked about was around the ethics of this stuff. It is, it is illegal in some places and unethical pretty much everywhere, for you to use other people’s voices without their express consent. I mean, that, there’s, there’s no short way of, there’s no easy way to say that.

John Wall – 14:40
Yeah, no, that’s, that’s the right PR image statement for us to put out there, particularly because…

Christopher Penn – 14:46
The technology right now does not prohibit you from doing it. There’s, there’s no technological way to say, “Oh, that’s not your voice.” So the technology will do it. So the, I guess, the world is relying on you to have functioning ethics and morals to not do things you shouldn’t.

John Wall – 15:09
Yeah. Yeah. All right. Well, that’s one of those things they talk about where it’s, we have things we choose to believe which may or may not be true, but we’d like the world to go in that direction.

Christopher Penn – 15:20
Exactly. The second tool that I have been playing with a lot is a tool called Suno. This is one of the many AI song composition tools that you give it a prompt, and it will help you create songs.

Now, the challenge with this tool is that it’s relatively limited in terms of what you can provide. You can either provide a 200-character general prompt, and they’ll try to come up with something, or if you hit Custom Mode, you can give up to 3000 characters of lyrics, plus 120 characters of text that you can then have it create content for you. In general, it does a really good job with instrumentals for background music. Like, if you wanted to have intro or outro music, it will do a really terrific job with that. In general, when it generates the lyrics, they’re terrible. They are.

John Wall – 16:16
Just like random word art, poetry word salad.

Christopher Penn – 16:19
It’s, it’s very trite and banal, and it’s like ChatGPT had a really bad day and was just like, “Do the minimum amount of work possible.”

John Wall – 16:32
Slacker AI larynx, I like it.

Christopher Penn – 16:34
Exactly. So if you want to use this tool to create songs, I strongly recommend that you have, you either write the lyrics yourself or you use a different language model to write the lyrics so that you can then just import them here and get a better-sounding song.

But what’s cool about it though, is it does generate pretty decent music. Let me go ahead and let’s go, in fact, let’s just create a new one. Let’s do go instrumental. How would you characterize in, in relatively short terms John, how would you characterize the intro music to Marketing over Coffee? The song is “Melody” by Funk Masters, but how would you characterize it? In fact, yeah, real quick here. So that’s the 2011 version we use for our show. We used the old seven version that was on the Podsafe Music Network at the time. But that is, that version is long gone.

John Wall – 17:41
Yeah. I think it’s, they’d call it like trans jazz, or there’s another subgenre in there. But it’s basically trippy jazz.

Christopher Penn – 17:55
Trippy jazz. What are the instruments in it?

John Wall – 17:59
Oh, that’s a good question. Well, we’ve got bass. I guess you just call that synth.

Christopher Penn – 18:11
Okay. What else?

John Wall – 18:14
I go with, because it’s electric guitar, but it’s not distorted, too.

Christopher Penn – 18:22
Okay, so clean electric guitar. What else?

John Wall – 18:26
Yeah. God, man. I have to go back to, like, I’ve never, like, dissected it. I’m just like, “Oh, it’s just playing in the background.” And not even there. Let me see here if I can pull out the rest of that.

Christopher Penn – 18:37
I’m pretty sure it’s a major chord. Major. It’s a major key. All right, let’s go ahead.

John Wall – 18:48
Yeah. Synthesize snare and synthesize clavier.

Christopher Penn – 18:59
Okay, that’s useful. So let’s go ahead and spawn those. So on with Suno, on the free version, you get basically enough credits to do 10 generations a day. And you, as you can see, just by hitting the Go button, it’s created two per prompt, so you burn through, basically, it’s effectively five uses a day.

If you pay for the paid version, which I’ve paid for the paid version because I find the software enormous, entertaining, then you get 3000 credits a month, which works out to like 300 songs a month, give or take, which is terrific.

The outputs from this, you get them in MP3, WAV, and video. They all do take time to render. But if you’re on the paid plan, they tend to render a little more quickly. Usually, they’re done like 2 and a half minutes on the paid plan, fully.

And if you do lyric versions of the videos, then you get an actual download. So this is an example. So this is one that I built from Ann Handley’s newsletter. She wrote a really great issue a couple of weeks ago. I fed the text of the newsletter into Gemini with a really long prompt for making, how to make lyrics from text, it’s like four-page prompt. And then it wrote the lyrics, and then this is what it came up with.

(Music plays)

Christopher Penn – 21:17
What do you think, John?

John Wall – 21:17
Yeah. Though this has that kind of early Taylor Swift, it’s a little bit. It’s a little too poppy and upbeat for Taylor Swift country, but it’s totally kind of hit that spot. I was surprised. The lyrics on that are definitely. But like you said, you have a crazily well-trained prompt that you’ve spent a lot of time refining to get that to be where those could be, right on the mark.

Christopher Penn – 21:38
Exactly. And there’s a reason it sounds like early Taylor Swift is because I had, I started with the lyrics from “You Belong With Me,” and I said, I asked the AI, “Don’t look, don’t use any of the original lyrics, but understand the writing style, diction and prose, rhyme structure, and all that stuff.”

There’s like 20 different components of writing style just for song lyrics. And I had the AI essentially extract out what those were as a set of rules and then turn that into, “Okay, now apply these rules to Ann’s newsletter.”

And that’s what it came up with, the camp was something that sounds very much like that. And then, of course, I asked the tool, “What do you know about that song? What’s the instrumentation? What’s the chord progression? What’s the, the timbre?” and all these, the aspects and that. And you get something that sounds very much like that. So our synth clavier thing has rendered. Let’s see, just give this a few seconds to see what this sounds like. I mean, that’s definitely something you would hear in an elevator somewhere.

John Wall – 22:53
It’s very, it’s close. It doesn’t have as much driving beat, so if we were to continue playing with it, it would need to get kicked up a little bit more. But it’s definitely in the same neighborhood.

Christopher Penn – 23:04
Well, that was without. We, we neglected to have the clavier and the synth and stuff in there. We only had a few things that initial prompt later on. We specified the synth drum kit, the synth snare, the synth clavier, and stuff. And that generation, actually you can see is a, it’s about twice as long because there’s more stuff to process for it. And let’s see what that one sounds like.

John Wall – 23:32
Oh yeah, that definitely. I don’t know, I had hallucinated the guitar, too, because there’s no clean guitar in the original one, but it totally fits in there. It does work. That, that is like right on the groove there.

Christopher Penn – 23:51
Exactly. So what you can do with this particular tool is a lot of, I think it’s most useful for backing track, backing music. One of the things that’s very popular, for example, if you’re in, when you’re doing online courses or when you’re doing a video podcast, if you’ve got a section of your show we’re doing stuff on screen and you just don’t want dead air, generate 10 to 20 themed tracks that sound like, maybe, maybe you have a known piece of music for the, for your intro, but you can add that in as an influence and say, “Generate stuff that’s kind of like that.” And that becomes your film music for anything. Anytime you’re, you’re essentially, showing how something’s made.

John Wall – 24:39
Yeah. And that’s great for a lot of stuff in audio and video. Where if you have a track that has a problem, if there’s some kind of weird, annoying whine or static or whatever, by throwing a super low music bed or noise bed underneath it, you can make that stuff vanish in the crowd.

Christopher Penn – 24:57
You can hide a multitude of sins that way.

John Wall – 25:00
Yeah, exactly.

Christopher Penn – 25:03
The last tool that we’ll talk about is video impersonation. And the reason we’re only going to talk about this is because it takes forever and a day for these tools to generate anything. And it takes quite some time. But what you can do with these things is you can essentially load the same types of audio into the system, train it on a video of you, and then have it render that and render it depending on what language you want in and things.

So this is one of those things where it’s uncanny, like you’re well into uncanny valley. So here’s an example of one.

(Video plays)

Christopher Penn – 26:10
So that’s it, having me speak Italian. Now, I did run that by some native speakers of Italian, and they’re like, “It’s not bad. Like the translation is good, the language is good, and it’s pronouncing the words properly.” It’s not, they said, “It’s not human, it doesn’t sound natural.” But if, again, if you’re using it as something like background or filler, it would pass a casual sniff test.

John Wall – 26:23
That’s, yeah, that. I don’t know. It just seems like video is a whole another level because to do sound and to do a single image is one thing. But then to have it actually, it just blows my mind how much processing juice has to go to that to do something at 30 frames a second.

Christopher Penn – 26:41
Exactly. Now, the video one is really tricky because that is, so that one has more protections. When you sign up for that service, you have to authenticate who you are. And if you’re creating an AI avatar, it then makes you turn on your webcam and makes you turn your head. It tells you what direction to turn your head in and stuff so that they. Because for obvious reasons, you could deep fake really well with this, otherwise.

John Wall – 27:06
Oh, yeah, that’s a good point. So that takes and puts some gates out, whereas if you don’t have access to that, you can’t just fake anybody.

Christopher Penn – 27:14
Exactly. There’s another version that actually uses a 3D scanner on your phone to do like a 3D map of your face to make sure you’re not holding up a photo or replaying video on a flat tablet.

John Wall – 27:28
Okay. Well, make sure you’re using a reputable vendor for that then before you’re giving up your digital head to somebody who’s going to have that in the future to get into your safe or whatever.

Christopher Penn – 27:38
Exactly. So those are some of the multimedia tools, I think, that have interesting applications for podcasting. We did cover tools like the script and Opus clips and stuff in episode one of the series, but these are some of the ones that I think are more on the edge. And the use cases for them are, some of them are pretty clear, like the backing tracks and instrumentation. Some of them are contextual. So, for example, if you happen to come down with, I don’t know, bird flu, having the predefined voice in the system, if you’ve got to get a show out, that might be an option as opposed to hacking your way through a show.

John Wall – 28:21
Yeah, that could be interesting. Have a backup benchmark.

Christopher Penn – 28:25
Exactly. But there are obviously the more sophisticated a tool gets, the more ethical gray zone you get because a lot of these things, you can see how they get misused pretty well.

John Wall – 28:41
Yeah. Well, and then there’s the whole, there’s all kinds of stuff that’s happened with music that we’ve already seen with people playing with this kind of stuff. I just saw a crazy thing where somebody took Dua Lipa’s “Houdini” and redid it as if Phil Collins had done it. And they were able to nail it. Now, it still had, there was a human element. They knew how to break down the Phil Collins sound, but it works. And yeah, and then the other interesting one too is this whole idea of artists who can recreate their earlier voices.

I mean, so many times we’ve had musical bands or soul performers who kind of over the years can’t sing the stuff that they used to sing. So what happens if they can just load a filter now and get themselves back? I mean, at least for that point, it could be for their own enrichment and extend their careers. But then there’s the whole question of, like, well, what if anybody else wants to use these tools? What happens then?

Christopher Penn – 29:36
Right. Well, I mean, there definitely are use cases for that, especially if you have a band or a musician that has a very distinctive sound that is very short-lived. So back in, was it the nineties? I think there was like a teen kids band, Hanson, and like those kids grew out of that voice print in like 2 years, but they still have all the studio masters.

John Wall – 30:02
So they could train and, sing as adults now, but still wind back the years to get the sound that they wanted to get.

Christopher Penn – 30:10
Exactly. The other ethical issues, and I think this is worth pointing out, is that in particular, like the sound generation capability for, in Eleven Labs to some degree and even Suno to some degree, does take work away from other people. So if you’re selling stock audio libraries, the Eleven Labs tool will generate pretty good candidates that, again, for most casual use cases is good enough that you don’t have to buy a stock library anymore.

John Wall – 30:41
Yeah. And that’s, it’s one thing if it’s, that suddenly brings some new tools to projects where they would never have the budget to do that in the first place. But, yeah, then there is the whole other ethical question of, like, okay, at what point should you be supporting an artist? And how does that all work?

Christopher Penn – 30:59
Right? Exactly. And frankly, there’s another interesting avenue as well where AI, this is the case. In most places on the planet, AI-generated content that is wholly generated by AI cannot be copyrighted because of a number of rulings that basically said only copyright can only be held by humans. And we should probably hang on. I think I have a lower third here for this. As a warning, neither of us are lawyers. Do not take legal advice from randos on the internet, please. But AI-generated content made by machine cannot be copyrighted.

Now, portions of stuff can be. So this is an interesting case with derivative works where in Ann Handley’s newsletter case, I used Ann’s newsletter, and I used AI to generate lyrics from it. You can tell that is clearly a derived work because it matches the text of her newsletter in that case.

And again, this is what my lawyer friends tell me. I am not a lawyer. The derivative, the derived lyrics retain Ann’s copyright because Ann wrote the original, and it’s clearly a derivative work. It’s not a transformed work, which in turn means that the AI song generated the music portion is public domain. You can’t hold copyright, but the lyrics portion and retains. So that song still has some aspects of copyright, which then means that I can’t go and legally put it up, like on Spotify because I’d be violating Ann’s intellectual property rights. I can’t play it on a, in full, on a podcast other than for, like, fair use reasons because it, and still has copyright to the lyrics and things.

So that’s kind of another interesting twist with some of these tools, depending on how you’re deriving the content, will depend on how the content inherits some forms of copyright. So if you were to take an episode of your podcast and have it, have all these tools summarize it by turning it to a song, if the original episode is fully human-made, like an episode of Marketing over Coffee, then the derived work portions of it, you would still retain the copyright to.

John Wall – 33:18
Yeah. It’s still your content that’s useful.

Christopher Penn – 33:22
Exactly. And on the flip side, and this is a part where, again, not a lawyer, but copyright, only copyright’s purpose is to protect your intellectual property rights, to say that no one else can use it. If you publish something that is public domain, you can still use it, you can still sell it. It’s just you can’t protect it if someone else wants to sell it. If you put the Mona Lisa on a T-shirt, you can sell that T-shirt, but the guy next door can make the exact same T-shirt, sell it, and you have no recourse because. But it does not prevent you from selling it.

So one of the interesting things to think about for intellectual property, for creators, is if you have these digital artifacts, you can actually make use of them and commercialize them. There’s nothing that will stop you from commercializing a piece of public domain content. Go on Amazon, you’ll see plenty of people reselling classic works that are long been in the public domain.

That means, though, that, for example, AI songs that I happen to create, I can legally go and put them up on Spotify or whatever. And I’ve actually got one, a track right now that’s waiting on approval to go into Spotify and Apple Music and stuff, machine-made. And I’ve disclosed it like, “Google Gemini is listed as the co-author on the song.”

But if you’re going to do this for your podcast and you’ve got these materials anyway, put them up and see what happens.

John Wall – 34:52
Yeah. Just float it out there and see if you get any traction with it.

Christopher Penn – 34:56
Exactly. And there’s any number of services. I’m using Amuse, there’s DistroKid, there’s all these different companies, but like I said, you can’t enforce the copyright on it. Someone else could make a copy and sell it under their name, and that would be, that’s legal with a public domain work unless the lyrics are yours, in which case you can enforce the copyright on the lyrics. They can copy the music, but they have to provide their own lyrics.

So that is sort of the ethical side of this stuff. There’s, there are definitely cases where you want to use a human musician. So, like, for theme music that I have on my YouTube channel, I hired a human musician, Ruby King, to write that, partly because I wanted to have a song that was thematically similar to my old show song, but I don’t have a license to use that song anymore, and partly because I want to be able to enforce copyright on the music itself. And so, hiring a human is the only way to go there.

John Wall – 35:52
Yeah, I can see that. That’s kind of the classic, if you’re making content that’s going to stay around forever, you want to make sure that’s clean.

Christopher Penn – 35:59
Yeah. Anything that’s going to be a part of a big campaign, anything where you would be unhappy if a competitor stole it, you definitely want to go fully human to protect those intellectual property rights.

John Wall – 36:12
All right.

Christopher Penn – 36:14
All right. Any final thoughts, John, on ways that you might be using some of these technologies now?

John Wall – 36:21
You know, unfortunately, like the first things that come to mind are horrible music ideas, like restoring all these artists’ albums to their former glory or whatever. But, yeah, it’s the hottest one for me. The idea of having foreign language just on, on hand like that to be able to at least get a passable copy in another language. That’s the most exciting and interesting to me. So after that, I’ll, I’ll leave the crooks to run with the rest of the stuff.

Christopher Penn – 36:51
The other thing that you can do is, this is kind of a fun one. One of the things that people forget about with these tools is that you can do some very creative things with them and have interesting applications. So, for example, taking Ann’s newsletter and turning it into a song, I was giving a workshop in Los Angeles, and I took the company’s About page from their website, put it through Gemini, put it through my songwriting prompt, and I turned it into a song for that organization.

I played it as the opening for the workshop, and to this day, people are still asking for copies of it. If you are making a big pitch, right, you’re an agency, you’re making a big pitch to a potential client, what an interesting takeaway that would be to summarize your pitch. Have Gemini turn it into lyrics and deliver the music video that goes with your pitch.

John Wall – 37:48
Yeah. Have a full multimedia and musical review together.

Christopher Penn – 37:53
Exactly. One of, one of the prompts that Suno does super well, Broadway musical solo as the music case. So you can take the contents of your pitch and say, “Okay…”

John Wall – 38:07
“We’re…”

Christopher Penn – 38:08
“…we’re flying out to Seattle to pitch T-Mobile. Here’s, here’s the concept of a pitch. Make a song.” And you can arrive with something that will definitely set you apart at least for a little while until everyone else catches on. Something that will set you apart.

John Wall – 38:26
Yeah. Get our full “Defying Gravity” moment there. That’d be…

Christopher Penn – 38:34
All right. That is going to do it for this week’s episode of So What? Katie should be back next week. And next week we are going to live rebuild KDGPT, like the Six Million Dollar Man. What was it? The, how did that go?

John Wall – 38:50
Better, stronger, faster?

Christopher Penn – 38:52
Better, stronger, faster. That’s right. So we’ll have a new version of KDGPT, and we’ll talk about why you need to regenerate your prompts and regenerate your custom GPTs because it is important that you do that for maintenance. So thanks for tuning in, and we will catch you on the next one.

Thanks for watching today. Be sure to subscribe to our show wherever you’re watching it. For more resources and to learn more, check out the Trust Insights podcast at TrustInsights.ai/tipodcast and a weekly email newsletter at TrustInsights.ai/newsletter. Got questions about what you saw in today’s episode? Join our free Analytics for Marketers Slack group at TrustInsights.ai/analyticsformarketers. See you next time.

Need help with your marketing AI and analytics?

Leave a Reply Cancel reply

Subscribe to our Weekly Newsletter

Pin It on Pinterest