AI in Podcasting – PCI 348
AI in Podcasting
Using AI for podcasting can enhance various aspects of the podcasting process, from content creation to enhanced search engine results. In today’s episode we discuss some ways AI can be utilized in podcasting.
Today’s Hosts: Todd Cochrane and Mike Dell Things you can do with AI- Help with writing show notes. Generate detailed show notes, summaries and more automatically.
- Suggestions for episode titles.
- Transcripts: Text-to-speech (TTS)
- Social media: Check out our new partner, Momento AI
- Clips: Endless content
- AI can even become the host
- Quickly write show notes
- Get fresh and useful ideas for titles and topics: using AI to brainstorm ideas is one of the most useful ways to refresh your podcast
- Sounds or reads like a machine created it
- Wrong information (it is using existing info, even if it’s incorrect)
- Plagiarism worries: Generated content always has an origin point
- Bad for SEO: Google will be able to detect AI at some point
Not mentioned in this episode, Planet Money from NPR is working on a series of AI created podcast episodes. Learn more about that here and here.
While AI can bring numerous benefits to podcasting it’s essential to maintain a balance and to preserve the human touch. AI tools should enhance the podcasting process as opposed to replacing the human creativity and authenticity that make podcasts unique.
The best place for support with any Blubrry product or service is our ticket system. Tickets give the whole team access rather than direct emails or calls.
Fill out our listener survey at surveys.blubrry.com/podcastinsider
- Hosting customers can schedule a one-on-one call with Todd or a tech checkup with Mike
- todd@blubrry.com and mike@blubrry.com
0:00 If used correctly. It's, it's a, a very interesting tool. And again, even though I had chat, g p t write a summary from an article, I still worry about the plagiarism situation here. Then you have to go back and, you know, did it take two sentences out of the article and put 'em in your summary. You're listening to Podcast Insider, a weekly podcast bringing insights, advice, and insider tips and tricks to help you start, grow and thrive through podcasting With the support of your team here at Blueberry Podcasting. Welcome. Let's dive in. This is Todd Cochrane, CEO o and co-founder of Blueberry. And I'm Mike Dell, the VP of Customer Relations here at Blueberry.
0:41 This week we're going to talk about the ways you can use this newfangled AI for your show and maybe why you'll want not to. So, Mike, ai, you know, it's, it's the buzzword of 2023 for sure. Yeah, certainly seems that way and it does a lot of neat stuff, but how should we be using it in podcasting at this point? You know, I think there's a, there's a whole bunch of companies that have come up with a whole bunch of tools and, uh, I've been playing with quite a lot of them and some of them do things better than others and, but I still think most of the solutions are still kinda half baked.
1:23 But I've been playing pretty hard using some tools just to see, you know, get some results and being very careful in editing and that type of stuff. But I think we can really rely on AI to help with things like show notes and maybe episode titles and potentially transcripts, social media clips, and even, well, I hope this never really happens, but folks are even using AI as, uh, to be a host. Yeah. I, I keep hear, I keep hearing about the one where they take Joe Rogan's voice and somehow with ai, make him interview people that are dead or, or aren't there, or whatever, and, and it, it sounds machine created, but you know, that technology's gonna get better.
2:11 I think for my particular situation, the way my show's laid out is I talk about multiple topics during the show and I usually have three or four lead topics, and then one of those lead topics is used as kind of my, my lead for my title for my episode. And what I've been doing is I've been taking the specific article that I've been talking about and I've been putting that into chat G B t I actually just, uh, copying the whole text right out of the article and then saying, create a short one paragraph summary and create a, you know, five word title and it does a good job summarizing those articles.
2:59 But then I get a little weirded out because you've almost at that point, you know, if you're, if you're following the rules, you need to make sure that you attribute, so what I've been doing now is I say summary created from an article and the link back to the article and then I edit it whatever chat g p t has come up with, I edit a little bit and add some personalizations, a lead in, a lead out so that I have, uh, texts there and you know, if I'm, I'm feeding it something, it seems to get me pretty good titles and, and pretty good descriptions and keywords and that kind of stuff just in just straight chat G B T. And I think from my use case, it, it works pretty good because again, I'm picking a specific topic, but I think a lot of podcasters are potentially gonna wanna run a portion of their transcript potentially through the AI to see if it can write a good show summary. I, you know, one thing this could really fix though, Mike, is the, the lack of show notes.
4:04 I mean, so many shows have so little show notes. Yeah. A lot of people just put a one liner in there and that's it. And you know, that doesn't help you for s e or anything and with chat G B T or any of the other tools out there, it can give you some ideas anyway. The thing, the thing that I worry about, uh, with this is people relying too much on it. So you, you know, you gotta feed it the right stuff to get their, the right output. And I think people have to be careful too, especially whatever they're using for source material. And you think people have to remember about AI at this point?
4:46 It is a language model. It is great for writing content that you provide the input for, you know, if, if I was, uh, going to say that I had an episode about, um, let's just use a historical, uh, figure. Let's say I was gonna do a podcast about Benjamin Franklin and I talked about the history of Benjamin Franklin, then I asked chat G B T to write a history of Benjamin Franklin. It may get it wrong, it may not have a good input, so you're gonna have to be very, very careful if you're just asking it to write a summary about a topic that you've covered. And I think there's high risk in that, whereas I've been in my utilization of it, of being able to use, you know, a given set of texts that was written by a reporter and being putting that into chat G B T and asking it to write a summary and then accrediting that summary back to the article.
5:54 I feel I'm meeting attribution requirements at the same time disclosing that the summary came from that particular article. Now I would, if I had handwritten that summary myself, which I used to do, I may not have attributed it because I did have it a link to the article later on, but it's because this is a bit of an expanded summary. I almost feel obligated now to actually do the attribution that way. Right, right. And uh, you know, that reminds me, there was a story recently about a lawyer who was ta doing a case and wanted to have some similar cases in his brief and he asked Chad g b t to look up similar cases to the case he was working on, and it came up with all these cases and come to find out none of them were real.
6:47 Whoa. So it just, you know, took stuff out of the air and made it up and, you know, he's in big trouble now cuz he turned in this, uh, this court brief and signed it . So you gotta be careful. Yeah, I think, I think so. You know, cuz it, it doesn't know the difference between a spoof article on the internet and another article on the internet that is truthful and, you know, they're, they're not that good at it yet. So part of the, part of the thing is you've gotta verify everything that it came up with. I mean, you and Rob did, uh, a demo where, you know, you asked for a biography and it was making up all kinds of crazy stuff.
7:32 Right. Some of the stuff that was in Rob Biles was actually about me. So it was, that was kind of the weird part. So I think it's one of those things, again, you have to, you know, use some common sense here. I think, you know, where I find the most useful, and actually we just partnered with a company out there called Memento. Uh, they're at Memento fm and what I like their tool for is after I published my episode, they grab the RSS feed and then ingest the audio. And then what they do is the, their, their platform analyzes the content from the actual transcript that it creates, and then it basically allows you then to make clips, shorts stuff can go on Twitter, Facebook, uh, YouTube, Instagram, uh, all the def different social destinations, and you can set a template or have a custom template and the text is not always perfect and the segments aren't always perfect, but you can kinda drag the segment pointer around a little bit to get it to the right start point, stop point or shorten it and make it longer and then kick out, uh, very, very quickly. Uh, social promotion.
8:49 Yeah. Like little video shorts and, and whatnot. And it comes up with, you know, stock video and stock pictures. Right. In the videos and they're formatted right for each of the platforms. It is slick. Yeah. And the thing I like about it too is I would've never done that otherwise. It's something I would've never, I didn't have time for, didn't have time to do shorts or clips or anything like that. But one thing that could be used is someone has someone that's assisting them with their show. They could say, Hey, go ahead and, uh, you know, find five clips and do a clip a day between now and the next episode or, you know, and, and basically keep the promotion going.
9:31 So I think that is, uh, where I see the, the best usage currently with, with Momentum. And I, I know there's a whole bunch of other tools out there for sure. But, uh, again, everything's not quite perfect at this point, and I think what's gonna happen that's gonna evolve very, very quickly and within a year, uh, most people are gonna be probably using the next model of chat, G B T and, you know, utilizing that tool for inexpensively to, to help in their production of their show. And I think maybe even production could come into the fact where let's say you're gonna do an interview with, uh, an economist and maybe they have a specific forte.
10:15 You could ask chat g p t to write 20 interview questions. A not for that person, but for an economist, and then do your research and edit those questions to your liking. But it might just, you know, help feed your brain and save you a little time. Yeah, that's one of the things I use chat G B T for is just getting ideas. You know, I'll put in a subject and have it write an article about the subject. I'm not gonna use that article, but I wanna see what it came up with and that may give me an idea of what I want to talk about. Right. And that, and same thing with interview questions and and whatnot. You know, of course I, I like to, when I do interviews, I like to just let the conversation go where it went, but, you know, sometimes you want key points, especially if you're talking to an expert in a particular field and all that.
11:05 So yeah, pretty, pretty cool stuff. There's, there's actually starting to be plugins out there now that will actually, you can, you can say, go over and look at this specific LinkedIn profile per se. Mm-hmm. . You know, and create the same, same type of thing. You, you give it, you point it at a destination to query and then it reads that page and then it can create a summary or questions or maybe a bio. So I can see if, you know, if you're reading a, if a guest has not given you their bio, uh, you know, maybe you can self-generate one off their LinkedIn profile. So it is, uh, I think these kinds of things just saves us a little time too.
11:47 Yeah. One of the things you've really, you know, one of the best practices is make sure you don't just cut paste that can get you in trouble. So, you know, you want to go in and edit it and, you know, sometimes it talks the way you wouldn't, so you wanna kind of customize it, you know, to, to sound like you wrote it. So when things for sure we're gonna want to, uh, you know, back check everything, but it is, it helps you with, you know, it is, I I agree it's going to make it much easier for your Right. Good show notes and again, ideas for titles and topics, and I always ask it like, write me an s e o title, you know, of five words for this particular set of text.
12:35 And, you know, the, the results have varied to be honest with you. Right. But it gets you in the right ballpark. Right. You know, sometimes, and you know, like I said, it's, it's not the cure-all for everything and you know, maybe to some people it's a little creepy even, but, uh, yeah, I, I think it, if if used correctly, it's, it's a, a very interesting tool. And again, it's, you know, just like the, even though I had chat g p t write a summary from an article, I still worry about the plagiarism situation here because I read that and then I go back. So I've read the article in my mind, I kind of know the type of summary I want.
13:18 Maybe I kind of directed to write that specific type of summary with some antidotes, but then you have to go back and, you know, did it take two sentences out of the article and put 'em in your summary. Um, that's where you have to be careful on this plagiarism. So that's why for me at least, I'm saying summary derive from, and then I give the attribution. Right? Yeah. Then one of the other things that, that I kind of hesitate about is SEO is Google or the other search engines at some point gonna be able to detect AI and somehow downgrade your, your page score.
14:02 I think if you're doing a copy and paste, maybe, but I also take another step, I take the summary that I get from the chat G B T, and I place that in Grammarly, and then I let Grammarly have a run on it and you know, and I look at it, say, could I, or would I have written what that summary or what that show note says, could I have done that by myself? You know, is it read like something I would've wrote? Mm-hmm. , I think that's the approach you have to take is when you get this output you, like you said Mike, you go back and you'll look at it and say, uh, I wouldn't have used that word, or maybe I did use a different sentence structure. You know, I'm not great in grammar, thank God for Grammarly and Grammarly is greatly improved, but even sometimes Grammarly jumps back and forth between, uh, two different things and it can't quite decide how something should be reordered. But I, I think that, uh, many of us are using, have been using Grammarly for years.
15:12 So I think some of this SEO stuff might have shown up just by upon that, but at this point, no, no, we'll see. Yeah, well Grammarly might be, uh, talking about using AI at some point. I'm. I'm sure they are to an extent already. Right. Because, uh, yeah, it's pretty amazing. Sometimes only thing I don't like is when I purposely misspell something it wants to change it, but, uh, you know, . So yeah, as far as AI stuff goes, uh, you know, I'm skeptical, but, uh, I think, I think overall it'll be a good thing for a lot of creators and could be a bad thing for some creators, you know, so if you could, uh, throw your, uh, audio in a something like that and have it just edit the thing for you, well that's gonna be tough for some audio editors to, uh, compete with.
16:09 I I think there is, you are already seeing it happening with some of the platforms out there. They're already starting to apply some being able to automatically take out, uh, os and ums and tighten up dead air. All that technology is not necessarily ai, but I think it's only gonna get improved over time. Yeah. We've been playing with another one, uh, another AI thing which was, is kind of a competitor for a phonic and actually used it on the episode with Aaron talking about artwork. He was using a, a microphone that wasn't quite up to par, which, you know, he sounded okay, uh, but ran it through this thing and it made a lot of difference.
16:52 So that's, you know, another, another way AI can be used and this company that, uh, is doing it, I think it's called in Sounds, and they use AI to figure out how to make the audio sound better. It doesn't edit the audio per se, it just levels it and makes sure that it, you know, some of the background is gone. It's, it's, it's pretty ingenious. I've been playing with it. Uh, again, like I said, we used it on Aaron's audio a couple episodes ago and I, the, the results were pretty amazing. I think overall it's gonna be pretty game changing over time. I think, uh, everyone's jobs are gonna be affected by it.
17:34 Production's gonna be affected by it in, I'm finding now though, that I'm taking more time to produce my episode because I'm going through some of these extra steps. So it doesn't necessarily gonna lighten your workload, but maybe the end product will be better. Right, yeah. Because like I said, you know, AI is just computers and they can only do what they can do for now, but uh, it's gonna get better as time goes on. So how are you using AI as a listening audience? We'd love to hear from you if you are. Yeah. Definitely. Drop us a note and, uh, let us know how you're using AI in your show and has it been game changing for you and, and maybe you have a tip for us.
18:20 You and I had a demo of somebody using AI for a lot of things. He was just doing it on his phone. We were at a party and that was just downright impressive. Uh, so the more you get skilled at it, the better it can get as well, cuz you know, you gotta ask it the right thing. All right, I guess it's gonna make it a wrap for our segment on ai. Thanks for joining us. Let's meet up next week and in the meantime, for more information to subscribe, share or follow, check out the show notes@podcastinsider.com to check out our latest suite of services and help Blueberry can help you leverage your podcast. Visit blueberry.com. That's blueberry without the ease.
Apple Podcasts
TuneIn
Deezer
Paypal