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 to approach podcast transcription automation to save time and reclaim your content archive. Discover how to transform your back catalog of audio into valuable text using accessible tools and artificial intelligence. Watch a step-by-step exploration as we build a system to download, transcribe, and identify different speakers in your podcast episodes. Understand the trade-offs between different methods, empowering you to choose the best path for your own podcast transcription automation and conquer those forgotten tasks.
Watch the video here:
Can’t see anything? Watch it on YouTube here.
In this episode you’ll learn:
- The benefits of podcast transcription
- How to automate podcast transcription with generative AI
- How to automate transcribing your podcast back catalog
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 – 00:00
So, well, happy Thursday, folks. This is “So What?”. The marketing analytics and insights live show from Trust Insights. I’m Chris, and John is here with me as well.
John Wall – 00:45
Yes, just had a case of beepilepsy. You know, nothing’s happening for an hour, and then suddenly, one minute before the show, all the things start ringing and beeping. So, under control.
Christopher Penn – 00:55
Exactly. So this week, we’re starting what we’re calling our summer makeover series, AKA maintaining all things that we don’t have enough time for or we’re unwilling to make time for the rest of the week, the rest of the year.
John Wall – 01:09
Yeah, exactly. This is like the list of forgotten tasks. Mine is like a mile long, so it’ll be fun to see what we pull out of this.
Christopher Penn – 01:16
I think it’s a cobbler’s kit, this. We’re making shoes for the kids. We’re the cobblers. We’re making shoes for the kids. That’s what the series is.
John Wall – 01:22
A bloody stump series.
Christopher Penn – 01:26
So, out of curiosity, John, before we get rolling with today’s topic, what is on your list of all the stuff that, you know, for our podcast, Marketing Over Coffee? What’s all the stuff that’s on your list of things that, like, oh, if only we had time for this?
John Wall – 01:43
You know, it’s one of the things is I have a couple of tools already running. Like I am getting video rendered from the audio file. You know, those get dumped. And there is some transcription stuff that we’ve done. And you know, all this stuff never goes the last mile. Like, I never actually get it up onto the website. So, in fact, there’s a service that I’m paying $20 a month for to render these videos that are going nowhere. So that’s right up on the top of the list, both of those. And then, of course, the newsletter is also a huge disaster. Like three months, no, not even three. I was looking at my own blog. It’s like a year behind. I basically just thrown in the towel on publishing so much of this stuff. So it’s.
John Wall – 02:23
Yeah, it’s a complete dumpster fire, really.
Christopher Penn – 02:27
Okay, well, today, we figured we’d talk about podcast transcription and ways to semi-automate this and possibly get to full automation. One of the challenges we have with Marketing Over Coffee is we are at what now? Let me take a look here at the most recent post.
John Wall – 02:46
Oh, 76, something like that.
Christopher Penn – 02:49
Yeah, we are now at episode 876. Really good transcription software didn’t come on the market till around, like, episode 700. So we have a very, very large back catalog of content that has no transcripts whatsoever. And even on the page themselves, like, this is this week’s post. We, we don’t have a transcript for it. So the first place we should probably start is what’s the bare minimum process that we could do for this podcast to get some transcription at least on there.
John Wall – 03:30
Yeah, at least up and on. I mean, you know, the file is in the feed, so the world knows it’s there. It’s just that, yeah, you know, there’s been no manual labor done on my side after that.
Christopher Penn – 03:41
Got it. So let’s, let’s see what we can do with this. The first thing, the most obvious thing would be, how do we get these files? The good news is that you have a really good naming convention, right? So we have, they essentially are all the same URL, give or take, right?
John Wall – 04:06
Yeah. And of course, it’s funny. Of course, this is the week that last week was a major screw up and it screwed up. But yeah, 858 of them are in the same format.
Christopher Penn – 04:17
Got it. So let’s take a look at this link. Now, there’s a bunch of different ways that we could get at all of these URLs, right? Because they all have the same format. The same format is traffic.libsyn.com, and we should probably put, make that a secure link. traffic.libsyn.com/marketing-over-coffee/moc876. Now, you could say to something like, you know, Microsoft Excel. Let’s, let’s, you know, split the URL apart, let’s do a drag and drop and just enumerate the field, right? So you could say 876, 875, 874 and stuff, and then concatenate and glue it all back together. That’d be the old-school way of doing it. Nothing wrong with that. However, that seems silly. That seems silly in the age of generative AI.
Christopher Penn – 05:06
So let’s just go to Google Gemini and say, I need to render the URLs for the last 100 episodes of my podcast. All I need is to decrement the episode number by one. Here is the starting URL. The next episode would logically be. And then we’re going to switch this to Gemini Flash, because we don’t need this world’s smartest model for this. Generate the next 100 episodes. Return your results in plain text, one episode per line. Now let’s see.
John Wall – 06:01
That’s quite the URL.
Christopher Penn – 06:03
It is. Oh, there we go. Yeah, it cleaned it up. Now we have the next 100 URLs. Let’s save this in a list. Look at that. It actually put it in markdown to redirect. That’s funny. To redirect from to Google. That’s annoying. I don’t like that. So I’m going to just put all that on a new line.
John Wall – 06:27
Wow. So they did that themselves. They’re grabbing the traffic.
Christopher Penn – 06:29
They did that themselves, which is annoying. I don’t like that.
John Wall – 06:33
That’s, you know, Skynet just keeps doing its thing.
Christopher Penn – 06:40
All right, let’s do a quick sort here and there. There we can see our URLs now. There are some episodes that are going to be a little bit off, but for the most part, that looks pretty good, right?
John Wall – 06:54
Yeah, yeah. Solid list.
Christopher Penn – 06:57
We’ll call this MOC Episodes. Now, here’s the question. I could, don’t want to, but I could go and download all these one by one just by putting the copy pasting the URL of the browser, right? That, that would be one way of doing it. However, that’s stupid. There’s absolutely no reason to do that. I could use some kind of browser extension, but I don’t want to add more software. I’m on a Mac, you know, which is what we’ve got here. So I’m going to go back to Gemini. And by the way, I’m using Gemini. You can use the AI tool of your choice, ChatGPT, Claw, Deep Seek, doesn’t matter. They’re all going to have about the same general idea of what to do. I’m going to say I’m using Mac OS 15. I have ZSH.
Christopher Penn – 07:46
I have a text file of 100 URLs of MP3s that I need to download. Wget and it is installed. Help me write a short command for Wget to download all 100 MP3s to my current directory wherever I run it. So we’re asking this thing like, hey, we know there’s already tools built into your computer. And if you’re on a Windows machine, you say, I’m running Windows 11. If you’re running a Linux computer, you already know how to do this. Because I don’t know any non-nerds that run Linux on their desktop computer. And it says, hey, great, here is the easiest way to do this. So let’s give this a try. I’m going to take my Marketing Over Coffee news list. Let’s go into this folder here, which there’s my episodes that we just downloaded all 100 of them.
Christopher Penn – 08:55
Going to start a new command line box and we’re going to paste what Gemini said. Oops, I need to change to MOC Episodes, which is the correct name. Wget C I MOC Episodes. Let’s see what happens. And look, magically, here comes lots and lots of podcast URLs.
John Wall – 09:23
If I just fire this up on hundreds of virtual machines, our downloads will go through the roof.
Christopher Penn – 09:31
I mean, yes, that’s technically true, but probably shouldn’t do that, right?
John Wall – 09:38
Speaking of ad fraud.
Christopher Penn – 09:40
Speaking of ad fraud. Now, here’s the next step. How do you currently do transcription, John?
John Wall – 09:49
The latest one, I have a, you know, we use Fireflies. I’ve got the other one, though, that I still pay for because I was too lazy to cancel it.
Christopher Penn – 09:59
Otter.
John Wall – 10:00
Otter. Yes, thank you.
Christopher Penn – 10:02
How much do those services cost? Do you know off the top of your head?
John Wall – 10:06
Yeah, no. This is the classic example of, you know, it shows up on my credit card and I don’t look at it. So yeah, I’m sure it costs more than it’s supposed to. I’m sure.
Christopher Penn – 10:14
Okay. There are ever so many tools out there that can do transcription and do transcription on your computer, which for, again, for Marketing Over Coffee. This is not a full-time business for us. We do not make enough income from the podcast to support as a full business. So anything we can do to cut costs would be a good thing. So it would seem to me that we would want to do the transcripts at the lowest possible cost because one of the problems with all the SaaS services is they either charge you per file per minute, or you’re given a budget of minutes or a budget of transcription hours, and then you have to keep paying. And the services, they used to be cheap. They’re not cheap anymore because everyone, their cousins adding AI features to it, which nobody asked for.
John Wall – 11:03
Right. The processing costs are not going down under any way, shape, or form.
Christopher Penn – 11:09
Exactly. So the current, very best transcription software that exists right now on the market that is free and open source is a model called Parakeet from Nvidia. Parakeet is their model. They trained on speech recognition. It is targeted for English. So if you’re trying to transcribe a language that is not English, your best bet is going to be OpenAI’s whisper. But what’s cool about Parakeet is that you can download it and run it on your computer, which means that you then don’t have to pay anybody to get good transcription. So if we go back to our folder where were. Let’s take a look at this. I’m going to pick that up. The episode at the top. I’m going to say Parakeet MLX. I’m going to paste. I’m going to say output format is Txt, because I don’t want.
Christopher Penn – 12:05
It can also do subtitles, which that itself is pretty cool. That episode is from episode 777. It’s about a 35-minute episode, I believe, give or take. And Parakeet is now chewing on this. It has read the file and in 15 seconds, it has produced a transcript. So let’s see if the transcript is usable. That’s okay.
John Wall – 12:35
It’s a wall of text, but it looks like it’s the right text, which is good.
Christopher Penn – 12:38
It is, and this is one of the challenges with our podcast is it is not. Almost all the tools that are on the market don’t have what’s called diarization. Diarization is when you say, I want to denote the speakers, like speaker one and speaker two, so that we get. Because they have different voices, they’re saying different things. We want to split this apart. So we have to figure out now, how could we diarize this text? How can we figure out what, who’s doing what? And there’s a bunch of different ways that you can do this. Some of today’s generative AI models can kind of do it, but they’re not particularly good at it. However, what we could do is we could say to AI, how do you do this, right? Say, I, I would like some help doing this, right?
John Wall – 13:28
Do some diary.
Christopher Penn – 13:30
So if we open up Google’s, if you go to Google Gemini, for those who are not familiar, this is the deep research feature, which is at the bottom here, which allows you to ask a research inquiry on pretty much anything and it will take between 10 and 30 minutes to go out and figure out how you know any, really any research task. So in the interest of time, I pre-baked. It’s like a good cooking show. I pre-baked this research project on how do we do local speech styrization on my Mac, just running on a Mac using Python, and it came out with a 24-page report. Here’s exactly how to do it with the libraries, some different solutions, the type of Python environment you’d want, the integration and how to deal with the very challenging case of how do you tell who’s speaking.
Christopher Penn – 14:29
It turns out, and this is something that I wish more people knew, someone already solved this problem. Like, actually many someones have already solved this problem. One of which is a Python library called Pianote that can do this. It can listen to file and say, give you timestamps back of saying, here is how to diarize this thing based on this. John, what would your next step be based on this 24-page research report?
John Wall – 15:00
Yeah, right. Well, so what this is, we have to feed this all into the LLM again and tell them to write the code for us.
Christopher Penn – 15:09
Pretty much. You say, this is the thing. So our next step would be, can we build a work plan? Can we build a product requirements document saying, I would like you to make me this thing? So what does that look like? If I go into our code here, which is called Audio Animal, which I, I couldn’t think of a better name for it. We say, let’s turn this into a full requirements document to make speech diarization software that can understand the speakers and split them apart. After that, you ask generative AI, I want you to build me a file-by-file work plan to bring this to life. And then after that, you put in your coding tool and say, follow the plan, write each file. And that’s the part that is so difficult to overstate. I’m not doing the coding.
Christopher Penn – 16:08
You’re not going to have me do the coding. Instead, AI is going to do the coding and the error checking and the validation and the cleanup and the syntax and all the stuff that good coders should do that mythically good coders do.
John Wall – 16:24
Not documentation, forward coders.
Christopher Penn – 16:27
Exactly. And if we did our job right, and I go into my live stream and I grab just that episode, I’m going to put it in here and let’s see if it will actually do what it’s supposed to do. And now it’s lit. It’s opened up the file and it’s starting to figure out who’s speaking. And there’s always at a minimum three speakers in every Marketing Over Coffee episode. There’s you, there is the guest or me, and then there is our British introduction.
John Wall – 17:10
The voiceover.
Christopher Penn – 17:11
Yes, the voiceover. Lovely. Exactly. When I first did this, I was like, why are there three speakers? And it’s like, oh, it’s because there’s a third person there.
John Wall – 17:22
Yeah, it does pick her out of the soundtrack.
Christopher Penn – 17:27
So a big part of this is under for is understanding why. Why are you getting the data you’re getting now? This annotation library split the file, actually behind the scenes to split the file into probably 600 pieces because every speaker’s turn is split up as its own miniature, separate audio file. What comes out now looks like this: Speaker 0, there’s Justine. Speaker 1, Speaker 2, there’s John. And it missed me on that first turn. That’s okay. And now throughout the rest of the file, it now no longer looks like a wall of text.
John Wall – 18:06
Right? Yeah. This is a lot more readable.
Christopher Penn – 18:09
Exactly. So if I was to go back to. Do we even know what episode was it? Yeah. No. Seven. Yeah. What date was that?
John Wall – 18:24
Would have to be just under two years ago.
Christopher Penn – 18:28
Okay. I wonder if I can find it in. In the search box. Yeah, there it is. “Can you spare $3,500?” And let’s just double check to make sure that is, in fact, the correct episode. Yep, there it is. So now in here, we could, if we wanted to, go straight from our markdown file, just do a bit of find, replace, call that intro. Speaker two is John Wall, and speaker one is me. We save that, copy it, paste it in. And now this episode. Let’s just make sure. Yep. Podcast hit update, and let’s take a look at the. At what it looks like. So we got to get rid of the asterisks, but otherwise, this episode now has a full transcript. So here’s the question for you, John. In terms of this process, let’s get rid of those asterisks because that’s markdown formatting.
Christopher Penn – 20:15
We don’t need it. If this was the process for getting a transcript onto the site, would this, would this be something that you would do?
John Wall – 20:29
No, I wouldn’t go through this much lifting.
Christopher Penn – 20:33
Okay.
John Wall – 20:34
I mean, it is definitely. It’s better. It’s, yeah, it’s not far off from me just cutting and pasting over to Otter. I throw it through there, and it’s not doing as good, and it takes longer to do the transcription. But it’s that manual lift of like, okay, here’s the final thing. Now go over and get it into WordPress and paste it into the right place.
Christopher Penn – 20:55
Got it. So the next step then would logically be, if it did itself, meaning it went into WordPress, grabbed the post, found the MP3, transcribed, and put it back and updated the post. Would that be something then that you’d say, like, yeah, I can. I can trouble myself to run that script once a week?
John Wall – 21:16
Yeah, right. That would be, you know, like, if it could. If it would go find it and paste it in there, then, yeah, running a script is totally doable.
Christopher Penn – 21:25
Okay, so how would we go about doing that?
John Wall – 21:30
We gotta have it just write some more code, right? And we just need to solve this like everything else with Python, pretty much.
Christopher Penn – 21:38
So we would start with research to say, hey, how do we do this? So fire up a Gemini Deep Research report. And again, remember, you can use this with pretty much any vendor. Claude has deep research now. OpenAI chat GPT has deep research. Perplexity has deep research. Brock has deep research. Everybody’s got deep research. Say, I guess we need to be able to download stuff from WordPress, store it in a database so that we can edit it locally, then get it transcribed from the same software. And it could be any software if you have a vendor you’re already paying for. I’m just going to use, I would just use mine because that makes more sense to not have to pay for it with 700 episodes.
John Wall – 22:22
And.
Christopher Penn – 22:22
Then once the edits are done, upload it back into WordPress. So again, we’d follow the same process. We did the research. Now we would have to go and say we need a product requirements document and ask some questions, have it build us the requirements document, have it build us the work plan, then have it go write the code. And what you should end up with is again, a piece of software that can go and do the thing. And in this case, the thing we’re not building in the transcription software because that, we already wrote that, why bother? Why would you reinvent the wheel? Again, in the requirements for the transcription software, one of the things I thought of was to say, well, make this have an API so that other software can talk to it.
Christopher Penn – 23:14
Because we did that, this can now say, okay, I’ll go find that API and I’ll go and do that. So before this show got started, I was essentially on the last mile of that. I’m running into some issues with security with WP Engine, which is where our website is hosted, so it does not run quite yet. And I’m not going to make everyone sit and watch me debug on the live stream. That’s just the definition of dead air. But it passes the tests that we set up to say, okay, can you download this? Can you edit this? Can you upload this? Because we always try to do what’s called test-driven development. It’s one of the magic words which says you write the tests first, and then you write the production code from the tests. That way you know the test, it all works.
Christopher Penn – 23:58
And so far on the back end, it passes all those tests. So it’s just a question of getting WP Engine to play nicely with us now. So that’s. If this existed, then would it be worthwhile going back and doing all 700 episodes that don’t have transcripts?
John Wall – 24:15
Yeah, yeah. I mean, if it’s just a matter of, especially because it’s one of those projects where you run it once and it’s done.
Christopher Penn – 24:21
Exactly. So this is the process of doing full podcast transcription automation. And the number one question someone’s going to ask is, can we do this without coding? The answer is not really.
John Wall – 24:41
It’s not. It’s no code, not no work.
Christopher Penn – 24:44
But even in a no-code solution, yes, you can talk to WordPress. So, for example, there is the very excellent software solution N8N, which we’ve talked about many times on the live stream. And N8N looks a lot like this and you might have. In fact, let’s see if we can do a little WordPress. There’s WordPress and we can say get many posts and let’s create a new authentication key here. We’ll call MOC HTTPs WW Marketing Over Coffee. And let me put this on another screen so I’m not blasting my password all over the internet, which nobody wants that. And connection test. Successfully. Successfully. So this will be. We’ll use the MOC. That one. We’ll say get many. Let’s get the last 99 and connect it up here. And it got 10. Thanks. That’s. Oh, you know why?
Christopher Penn – 25:58
Because the RSS feed only has 10 items in it. I’ll bet you that’s what it’s doing. But yeah, so there is the content of the posts and then from there, we would need to have to start building the actual automation of can we get this thing to find the MP3 from the body text? So let’s see if we can extract text string. What is the widget? Is it a data transformation? I think it is data transformation. Let’s see. Filter, limit, split. It’s been a hot mess. I’ve done this. I want to say I think it’s edit set, and then we need to map in the rendered content, and then we need to write JavaScript to find the MP3 link in here. I have no idea how to do that. So that’s Gemini.
Christopher Penn – 27:18
I need a JavaScript expression for N8N to extract the MP3 URL from this JSON blob and we’ll go back in here, we’ll copy this whole thing. Actually, you know what we could do is just take a screenshot of this and behind the scenes is having to remember what the heck N8N is and then pull out the JavaScript so this theoretically. Look there. Now it can now extract out this and we’re going to call this field MP3, and let’s go back to our canvas, hit run. And so now it should have a list of 10 MP3s up. There’s one post that doesn’t have one. And now we have to figure out how the heck are we going to download those files and send them to the transcription software. If you are using a commercial service, you can do an HTTP request.
Christopher Penn – 28:50
You can say, I want to send this to. I’m going to do a post and let’s see if this works. HTTP 1/54321 and let’s see if we can send the MP3 and send body JSON using the fields below MP3, and let’s see what happens. The, the software says, hey, you sent me this thing, but that’s not an MP3 file. So what I would have to do is either build something in N8N to download the MP3 file and send it on, which would be a pain in the butt, or revise my software to do that. So even in a case where you’re talking about no code, there’s still a lot of work to set this thing up.
John Wall – 29:59
Right?
Christopher Penn – 30:02
But if we can get it working, if my intent is to get the Python version working, because I find that’s more durable once it. Because it has things like unit testing and error checking and stuff, which N8N does too. But I, being a frail human, forget to do that and as a result, I, my code is full of errors.
John Wall – 30:23
That’s interesting. It’s something to see inside of N8N because I haven’t dug around here at all, because it’s interesting how there’s so many apps that it has on the list and you think with all those pick lists you’ve got all the flexibility you need. But as you can see here, it’s kind of like as you keep digging down further and further, you run into stuff as you get down to the ground level of it.
Christopher Penn – 30:44
One of the things I found with N8N is that there is a point after which you start having to go through such convoluted workflows that you’re better off just writing code at that point. You’re better off saying, you know what, just make the Python code for me, because at least I know that Python will run it.
John Wall – 31:03
Yeah, that’s interesting. Well, how much work is it to get started at the top then with Python? I don’t know, it seems like it’s not that much work. I mean, you’re just telling the tool, hey, this is in X, you know, it’s not as if those pre-made menus are saving you all that time.
Christopher Penn – 31:18
Yeah, it’s again, with today’s AI tools, I would call it relatively minimal. You have to know how to run a Python script, which you know. So you have to know where the terminal is and be able to open up the command line and type in the command to start your app. But besides that, almost everything else, just generative AI can handle because writing code is one of the things it is absolutely best at. The tool that I suggest people use these days is one called Klein. This fits inside Visual Studio code. It is free to use because you have to plug in your own AI services to it. In the example of Visual Studio code here, when I turn on Klein, down here is where I choose what AI provider I want to use. And so all, almost all these cost money.
Christopher Penn – 32:12
So this is where the money part comes in. You have to figure out which provider you want to use and what you’re willing to pay because yes, you may have like, you know, a ChatGPT subscription. That’s not what’s being used here. All these things use the APIs, and that is essentially a pay-per-word basis. Now, some of these tools are pretty decently priced. So, for example, Gemini Flash, $0.15 per million tokens in, $0.60 per million tokens out. And you can see for this project so far, to get the WordPress posting thing used, I’ve put up half a million tokens up, put 7,000 down, and the total cost has been 1.5 cents. So this is not a bank breaker by any means, but it is still something that will cost you money.
Christopher Penn – 33:02
However, once you build the Python code and it runs, the software is yours. You pay nothing for it anymore. So with, like the transcription software that’s done. It is done. It runs on my laptop. I pay nothing for it ever again because other than the electricity to operate my laptop.
John Wall – 33:19
Yeah, right, that you, it’s. And again, you don’t have to worry about updates or anything breaking that.
Christopher Penn – 33:25
Exactly. And you have to worry about like random new AI features being added to the software because you’re in charge of it. But also, if you think about on a cost basis, so what does, I don’t even remember what the pricing is on services these days. So now these, this is one set of tools. The, the basic. If I wanted to transcribe this much audio because, see, this is 1200 minutes a month for the 6,000 minutes a month at $20 a month. If I’ve spent 92 cents on API calls to generative AI, compared to just the monthly cost of a single month for the actual service, that’s. That, to me, is a pretty big savings.
John Wall – 34:10
Yeah, no, that’s all good. And then, yeah, you’re just like, you’re. It also only burns it when you run it. You’re not, you know, as I’m doing, paying for monthly services that you’re not even using. You’ve got all gym memberships for me.
Christopher Penn – 34:23
Exactly, exactly. So this is how. This is how to think about.
John Wall – 34:30
Getting.
Christopher Penn – 34:31
To a point where we can start commissioning our own software. And in this case, for podcast transcription automation. As long as we can define the problem and talk about it step by step with a generative AI tool, we can say like, hey, here’s what I want to do. Help me understand the process for doing this and any of the tools, ChatGPT, Gemini, Claude, whatever, will say, well, if you want to download stuff from your WordPress site and things like that, you probably will need to do this. If you want to transcribe it, you need to do this. And then from there, you can say, okay, keep breaking every task down into its pieces until you can say, well, how do I build software to handle this piece?
Christopher Penn – 35:07
And then like Lego blocks, you just start putting them all back together to say, okay, now I’ve got my Lego blocks for this or that, the other thing. And eventually what you end up with is a software suite that, yes, you have to maintain, but you also in, in the SaaS, in the world of SaaS, where every, we’re paying $15 a month for like 85 pieces of software. And we’re wondering why, you know, our, our Amex bill at the end of the month is $10,000 a month. Every month you start going, oh, yeah, I’m paying for an awful lot of software that. That is like gym memberships.
John Wall – 35:39
Yeah, exactly. And this is at a point I can. This workflow I could do. You know, I could have the. The file ready to go because it would be done by the time I’m ready to finally hit post on WordPress. You know, I could throw the file in as soon as it’s ready and then go finish up the posting process. And then, yeah, last thing is just cut and paste at the end. So, yeah, this is actually more doable.
Christopher Penn – 36:01
The other thing is that for almost every syncing service, there is like Google Drive, OneDrive, you know, Dropbox, etc. They all have URLs on your computer. So you could, if you wanted to put this, the transcription software, say in a Google Drive folder, right? It will still run on your computer, but if you were to say have a podcast folder on your, that was shared in your organization, one person from marketing could drop the MP3 into Google Drive and another person who’s got the software running on their computer, you know, it would just check once an hour, transcribe and just put the transcripts in an output folder.
Christopher Penn – 36:41
And now you’d be using the syncing software that you’re already paying for and the workspace software you’re already paying for and just having additional tools watching those folders and when something new appears, it just runs and you end up with the transcript.
John Wall – 36:58
Yeah, that’s pretty cool. Modern batch job, but using somebody else’s infrastructure. That’s killer.
Christopher Penn – 37:03
Exactly. And again, not paying for it. So if you had say a team of 10 content marketers, you wouldn’t want to have the software necessarily running on all 10 people’s computers because that’s just ridiculous. You would have, maybe it would have a server and then it would, that server would have the various pieces of code and then just, people run it. People just come in, drop off their stuff and come back, you know, in however many minutes or whatever to pick up their stuff. And it is, it’s almost kind of like the old-fashioned corporate mail rooms where you just like drop things off, pick things up and stuff a couple times a day.
Christopher Penn – 37:41
But it would save, you know, so for us this saves, you know, Marketing Over Coffee quite a bit of money, like hundreds of dollars because we’re not gonna have to pay to transcribe all those back posts. But that’s for literally a two-man shop. I could see it an organization of maybe 100 people, maybe in a thousand people. This could add up to savings very quickly because you’re not paying for 20, 30, 40 subscriptions to the software.
John Wall – 38:08
Yeah, no, that just makes sense. It’s one of the ways to just like get those unnecessary expenses out of the mix.
Christopher Penn – 38:15
So I, I guess on the next show we’re gonna have to figure out how to talk up to WordPress. Hopefully I can get that code straightened out between then and now. And if we can get the code straightened out, then all those other updates and things we can start saying, well, how could we wire together automations so that when you get your video, snip it up right. It can get, it can get put somewhere. Whether it’s because almost every tool in the world has an API and is, if you can have AI right to your API, then you can take all that stuff off your plate.
John Wall – 38:55
Yeah, it would just happen and get done. That’s exactly where we want to be. Have a power tool that makes the pain go away.
Christopher Penn – 39:02
The same is true for the Marketing Over Coffee newsletter. So we’ve talked about revamping in the past. Maybe we do a makeover. We say it’s just going to cobble together the newsletter for us. So we can pull as we saw with N8N, we can pull the last eight posts or however long. So if it just runs and pulls the last 10 posts, figure out which ones are in the past 30 days and makes a summary version of the newsletter. Maybe that’s an automation that runs and then it gets posted somewhere for you to actually hit send on. Particularly if there’s pre-written ad copy that needs to go in it. Do you have like ad copy?
John Wall – 39:37
Yeah, there’s, you know, we’ve used past clips and you could just put a slug in there that gets changed. But yeah, like that, you know, it would save a lot of time if it was just. I could go in and that ugly rough draft is done and I can just kind of do the intro and tweak a few things if I want a different voice on the things. But you know, all that, like cutting and pasting from every show, that all goes away.
Christopher Penn – 40:00
Yeah. Particularly if the transcripts are already done on the site. Then the AI can read the whole transcript and say, okay, this is what you know, here’s the one-paragraph summary of this episode and what it’s about. I think that would probably be. So that’ll be in a future makeover.
John Wall – 40:19
Yeah, I’ll have to put some time against that too because I have kind of my own process of how I do that. But I would have to codify that so that we can give the right instructions to get as close to what comes out of the thing manually.
Christopher Penn – 40:31
Exactly. So next week’s episode will be on leveraging these transcripts to turn them into bonus content. And that can be one of the pieces of bonus content. So, John, have your scripts ready, your process documented for next week’s show and we will wire that up.
John Wall – 40:47
That sounds good. Yeah, we’ll look forward to that. I gotta make sure it’s on my list so I get it done for Thursday.
Christopher Penn – 40:52
Exactly. That’s all gonna do it for this week’s episode. Folks, apologies is a slightly more technical episode because as you saw the, this is not a no-code solution. This is a, if you want it to be durable, it is. You’re gonna need to get your hands into the guts of the machine. But once you do, you save yourself an awful lot of time and an awful lot of money. So we’ll see you next week for more bonus how to turn podcast content into bonus content. See 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 TI podcast and a weekly email newsletter at TrustInsights.AI newsletter. Got questions about what you saw in today’s episode?
Christopher Penn – 41:44
Join our free analytics for Marketers Slack Group at TrustInsights.AI analytics for marketers. See you next time.
|
Need help with your marketing AI and analytics? |
You might also enjoy:
|
|
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday! |
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new episodes every Wednesday. |
Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.
Newsletter
Podcast
