March 6, 2024 • 47 mins
Transcripts have long been promoted as ways to make your podcast more findable and accessible. But it's only recently that podcast transcripts have become actually useful!
The post How to Use Podcast Transcripts first appeared on The Audacity to Podcast®.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
How to use podcast transcripts.
(00:09):
Thank you for joining me for the Audacity podcast.
I'm Daniel J. Lewis.
Transcripts have long been promoted as ways to make your podcast more findable and accessible,
but it's only recently that podcast transcripts have become actually useful.
I've wanted to talk about transcripts for years and I've talked about them in other
(00:29):
places and in social channels and such, but never here on The Audacity to Podcast.
So it's finally time to get to it.
And the timing of this is great because although I didn't plan the timing of these things
to line up this way, I just released a new feature for Podgagement that transcribes the
voicemail you receive from your audience through Podgagement.
(00:52):
If you want to send me a voice message, go to castfeedback.com/audacity.
You can ask me a question that I can address in a future episode, or you can just thank
me for the podcast.
If you want to just try out the system and see how it works from the listener's side,
that's castfeedback.com/audacity.
And then when you send a voice message through that system, you could also write your message
(01:14):
too if you wanted to, or you can attach a written message to your voicemail, which is
great for if you have anything that's hard to pronounce, the spelling of something is
ambiguous or you mention any URLs, it's great to be able to include that with the written message.
But when you send that, then I will receive a notification that includes the transcript
of your recorded voice.
(01:36):
So then I can very easily see what you're talking about, know how to categorize it for
my own podcast, or if I'm on the Podgagement site, I can easily filter through all of my
feedback based on what was spoken in the voicemails I received from you and other audience members.
So I'd love for you to try this over at Podgagement.com for the automatic transcriptions of voicemails
(01:59):
from your audience.
It's really cool.
It's a great way to engage with your audience.
That's over at Podgagement.com.
Please try it out.
The other thing that makes this great timing is the release of iOS 17.4 which brings podcast
transcripts to Apple podcasts and maybe even more exciting using the podcasting
(02:19):
2.0 spec. Huge thing there. I'll get more into that a little bit later in this
episode and also there's a separate episode of the future of podcasting where
Dave Jackson and I talked about the implications of all of that and I know I
addressed it briefly in my previous episode about the top features of
podcasting 2.0. Now if you want the links to these things that I mentioned then
(02:41):
then please go to the notes at theaudacitytopodcast.com/transcripts.
Little disclosure up front, several of the links on my site and things that I talk about
are affiliate links so I earn from qualifying purchases only through my links.
I am not currently sponsored by anyone so I don't earn money simply by talking about something.
(03:02):
I earn money only through some of my links, but I hope you know by now that I recommend
things I truly believe in, regardless of earnings. And in fact, some of the things I recommend
on the site don't earn me anything. So check out the notes, follow along if you'd like
a tap or swipe away or at theaudacitytopodcast.com/transcripts.
(03:23):
So how do you use podcast transcripts? Step 1. Don't believe the myths about podcast
transcripts. For many years, people have been promoting transcripts for the wrong reasons.
I've also been pitched by many transcription companies pushing the wrong approach.
While there are elements of truth to these two common myths I'm about to share, the
(03:46):
ways people often promote and understand these two ideas are where the information becomes myth.
Or could we say myth information?
Right?
Right?
Oh, come on, come on, please tell me that made you chuckle at least a little bit.
Send me a voice message through castfeedback.com/audacity to let me know what you thought of that.
(04:11):
Maybe we should just call it from now on, this is myth information, which is very hard
to say without saying it with a complete lisp.
Anyway, myth number one.
Transcripts are good for podcast SEO.
No they aren't.
People have often said that you should publish transcripts on your website for your podcast
to make your podcast findable through search engine optimization or SEO.
(04:35):
But that's not the full truth.
Yes, transcripts are better than nothing, or a mere paragraph or a list of topics through
bullet points, but transcripts result in a lot of words and actually very little content,
resulting in very low quality writing.
Which is probably even worse than an AI could create, or worse than what you might get from
(04:59):
someone who's not a native of the language. For example, here's a transcript of a two
co-host podcast that I've made up. This does not come from or it's even inspired
by anyone's podcast. So if this sounds like it's exactly your podcast, I did not listen
to it to make this fictional transcript. But it might sound something like this. And you
(05:20):
can read this in the notes if you want, but I'll read it for you.
Jack, welcome to our podcast where we talk about things to help you do stuff.
I'm Jack.
Jill, and I'm Jill.
Jack, before we get into this topic, how are you doing Jill?
Jill, the weather is sunny today so I'm great.
Jack, great!
Jill, yeah, what about you?
(05:41):
Jack, I'm fine but I don't know what the weather is but…
Jill, that's sad.
Jack, what do you mean?
Jill, I mean, you're saying that you haven't gone outside or even looked out a window in
a while.
"Yeah, but it's okay, Jill."
"So anyway…"
"Jack."
"Yes, let's jump into this week's topic."
Now note, with this fictional example, there are a lot of words.
(06:04):
But what did you actually learn from this?
What value did you get from this?
Probably nothing.
That kind of back and forth is fine in conversation, so I'm not saying don't have conversations
like this.
Sometimes even okay to be in a podcast, in appropriate places, and if done well.
(06:24):
But in writing, it becomes basically worthless.
Compare how that same "information" could be written in a way that's actually turning
that content into something valuable.
And I quote.
(06:45):
That's right.
Nothing.
That whole exchange could be edited completely out of the written content because there's
little to no value in reading that.
There might be some value in hearing that, and certainly in having that conversation,
maybe, but it's not really valuable to read that.
(07:08):
Search engines prioritize high-quality content that is readable.
But unless you're a professional speaker performing a refined monologue, transcripts
are neither readable nor high quality.
So no, transcripts are not the magic bullet to making your podcast perform well for SEO,
especially if you're relegating transcripts to a downloadable file or a webpage separate
(07:32):
from your podcast episode.
That's just not really helping you out all that well.
There is an aspect of truth, and certainly transcripts are better than nothing, but they're
not the magic bullet that everyone seems to be saying, "Yes, you should do transcripts
for SEO.
That's the reason you should do it, because it makes your podcast findable."
No, that's myth information.
(07:53):
So hard to say that without lisping overly.
Myth number two, transcripts make your podcast accessible.
No, not really.
It's true that people with hearing impairments still consume podcasts.
Thus, transcripts have been lauded as the accessibility solution all podcasts need.
But the problem is that most published transcripts are still difficult to read.
(08:17):
Like that example I shared with you.
That's not really easy to read.
Or worse, those transcripts are buried in a link or downloadable file that could actually
be even less accessible.
It's like putting up a billboard with a QR code that drivers can scan with their phones
to watch a video to learn why they shouldn't be distracted while driving.
(08:40):
It just defeats the purpose of it.
Also, good accessibility doesn't help only the impaired.
It can also help memorability or engagement for the rest of your audience.
For example, a word or URL that might be ambiguous when spoken, such as 2, 2, 2, and 2, or let's
Let's throw in a bonus one, 2, and that would be T0 vs. Two vs. The number 2 vs. T00 vs.
(09:09):
Let's throw in the Roman numeral 2, which would be two capital I's.
Which one are you actually saying?
Well, if there's a transcript that goes along with the podcast, someone could glance
at that transcript if it's properly displayed, more on that in a moment, and they could see
exactly which one is displayed.
Did you know I actually have the audacity number2podcast.com just in case someone somewhere
(09:35):
out there automatically transcribes me or writes my URL the incorrect way with a number
2 in it instead of the word T-O in it because it is a little ambiguous.
Yes, but that's a small investment of $10-$15 a year to make sure that I have that domain
so if someone gets it wrong it still takes them to the right place.
(09:56):
But that kind of ambiguity gets even worse if you've made up words.
Like podgagement is a completely made up word that I have the trademark for.
And some people might be wondering, "Wait, did he say podgagement or patgagement?"
And don't worry, yes I did register patgagement.com just in case someone mishears it or mistypes it.
(10:19):
But here, properly written show notes or an article can make the information far more
memorable and actionable than a giant transcription page.
But it's only properly formatted and properly published transcripts that make your podcast accessible.
And more on that in a moment.
But simply publishing a transcript doesn't instantly make your podcast accessible.
(10:43):
It's how you do it that makes the difference.
Those are the two myths I hear and see most often.
Now as you've noticed, there is an element of truth to them, but most of the time, the
way that people are promoting and understanding these things makes them, I just gotta say
it again, makes them myth information.
Oh, I'm way too entertained by that.
(11:05):
Step 2, and this one's optional, use transcripts with podcast production tools.
You can use the power of transcripts immediately after you record your episodes, even if you
never publish those transcripts.
While you could pay for someone to transcribe your podcast for you, AI tools have gotten
really good at making fairly accurate transcripts. And some of them can even listen while you
(11:30):
are recording and give you the transcript immediately after you finish recording.
Yes, AI has become the ubiquitous tool here and really for all kinds of content creators.
I've joked before in multiple places the line from It's a Wonderful Life, a classic Christmas
movie you gotta watch. I watch it every year. I've started crying more frequently now when
I watch it. But there's the classic line in there and here's my version of it, modernized.
(11:54):
"Look daddy, teacher says every time a bell rings, an AI-powered app launches on AppSumo."
Isn't that the truth?
While I'm not a fan of some of these AI apps that are coming out and how some people are
leaning too heavily on AI, I am a big fan of using AI on content you made instead of
(12:15):
using it for making your content for you.
The way AI-based tools work with your audio or video podcast is by first transcribing
your content. That's where the transcripts come into play here. Then, you can use the
AI to do impressive and time-saving things with that transcript. Like summarizing your
(12:35):
episode, suggesting episode titles, pointing out the best parts of your episode, or even
helping you better edit your recording to find the stuff that wasn't of much value
or find the stuff that you want to pull out and use in particular places.
There are a bunch of tools out there, and like I can easily joke, it seems like a new
AI powered tool is launching every day, throughout the day, on some of these startup sites.
(12:58):
But here are my current favorite transcription based tools to help when your episode is still
in this early production stage, even right after you've recorded.
It works great for audio and video.
And in fact, I will be using Descript to not edit the audio of this episode, but to pull
out some video clips of this episode.
(13:19):
I'll talk more about that idea in a moment.
You could also use Adobe Podcast for editing your podcast episode.
That works only on audio currently.
Or Adobe Premiere Pro works on video and you can actually even edit audio inside of Adobe
Premiere Pro because it has a transcription based editing tool built into it.
And Adobe Audition...
(13:41):
Someday maybe sort of?
Audition seems to follow behind Premiere Pro by maybe about a year on some of the features.
So I expect to see some of this cool stuff that Adobe is doing like with Adobe Podcast
and Premiere Pro.
I expect to see that coming into Audition and it's already in there in some places actually.
And the other thing is my favorite podcast editing app, Hindenburg.
(14:04):
I get paid nothing for recommending them.
They don't have an affiliate program.
I want them to have one.
Still, it's my top recommendation for an audio editing app for podcasting because it
has so many great things and especially with version 2, it now has transcription based
editing and you can export that transcript and use it for all kinds of cool things.
So each of these tools that I mentioned, and this could be an ever-growing list, I'm not
(14:28):
going to try and keep this list up to date because AI tools come and go, but these are
the most popular, the ones I expect to stay around. They're the most professional tools
as well. Each of these tools can make editing your recording as easy as editing a text document.
So that's why you might frequently see me recommend, especially Descript, recommend
(14:49):
that as a solution for someone just getting into podcasting who knows nothing about audio
editing. But besides just editing the audio or the video with the transcript, these apps
can also export the transcript for your finished production that you can then use in some of
the other steps I'm going to share in this episode.
(15:10):
Even if you don't use any transcription based tools, you can still use a transcript
of your recording to help you find places to edit, excerpts for sharing, and reminders
of your content, like when you're writing your notes or writing those posts that are
going to promote your episode, or maybe sending something to your guest, you can look back
over the transcript to remember, "Oh yeah, that's what we talked about," or "Oh yeah,
(15:32):
I need to make sure to include that URL in the notes for this episode because that came
to me just spontaneously while recording this episode.
I need to make sure that's included in the notes."
It can really help you do some of those things much faster simply by having the transcript,
even if you don't publish it.
But I still recommend that you do publish your transcripts.
(15:53):
More on that in just a moment, but if you want the links to these tools that I mentioned,
get them in the notes for this episode a tap or swipe away or at theaudacitytopodcast.com/transcripts.
Step 3 to using podcast transcripts.
Make properly formatted podcast transcripts.
In order to be actually useful, a transcript needs to have more data with it.
(16:16):
At a minimum, transcripts need timestamps for when lines or even words were spoken.
That's what enables closed captioning to work.
Closed captioning is not just a giant block of text in a document somewhere.
It is timestamped text.
So the system that's playing the video knows exactly when to display what lines of text.
(16:38):
That's the same way it works in a properly formatted podcast transcript.
Transcripts can also include the speaker's names, formatting, and more. The most common
transcript formats are VTT and SRT. But VTT, and specifically WebVTT, is the superior format
(16:58):
and what I recommend now because of some cooler things that you can do with it. In its vanilla
state VTT and SRT will actually look similar, but WebVTT allows you to do a whole lot more
on top of the normal timestamped transcript. If you look in the notes over at theaudacitytopodcast.com/transcripts
(17:20):
you'll get to see a little example of what one of these VTT or SRT transcript files will
look like. Where you have a time code range, so it's two timestamps, start and end, one
line of text underneath it, and then another timestamp range, and then another line of
text. Sometimes you might see multiple lines. You might even see some formatting in there if it's
(17:43):
WebVTT. But that's what it is. It's the transcript broken down into lines, sometimes maybe even
individual words, with timestamps for that information. And then WebVTT can add a whole
new layer on top of that with additional features and formatting and a lot of other really cool
stuff. So that's why I think VTT is the superior format and what I'll be recommending and switching
(18:08):
to from now on. But don't let the format of these transcripts scare you off because you can get
these properly formatted transcripts from the transcription-based editing tools I shared
previously, or you can generate them through AI or third-party help. And remember the most
important thing here is to get the transcripts in the right format with those timestamps in that
(18:32):
actual SRT or VTT format. A big block of text in a PDF or text document is not going to be as useful
as the structured data like you get with SRT or VTT and that structure is necessary for all of the
cool magic that can happen in podcast apps now thanks to Podcasting 2.0. Easy access to artificial
(18:56):
intelligence or AI tools has made creating transcripts a whole lot faster, easier, and
cheaper. The editing tools I mentioned all use AI to generate their transcripts. Or you could use
other tools and services too, like Cast Magic, CapShow, Affonic, Buzzsprout's co-host AI add-on,
Transistor now has an AI-based transcript tool built in, Otter, Rev, and other tools here and
(19:21):
there that can help you generate those transcripts and some of these tools can do more things with
those transcripts. This list also is not going to be up to date forever because a lot of other
tools come and go with these cool transcript based features that have AI built in. But these are some
of the best ones I think right now. I like the more modern AI based transcription tools because
(19:45):
they understand language better. They're called large language models, or LLMs. And
because of that, they're able to make transcripts that actually read well and are usually quite
accurate, even with pretty good punctuation too. Compare that to transcripts from only
a few years ago that transcribed what they thought they heard, even if it didn't make
(20:10):
You can look up on YouTube stuff like "caption fail" or "transcription fail" and I used
to do this and make fun of this myself with my comedy podcast where we would use Google
Voice which would transcribe the voicemail, but the transcripts were just what it thought
it heard even if it didn't make sense.
And so we could sometimes read those transcripts instead of playing the voicemail and reading
(20:34):
it just did not make sense at all.
It would insert a random word here and there just because it thought that's the word it heard.
But with these AI systems that know how to work with language, they're much better at
understanding this is probably actually what you meant and it's pretty close to the sound
and this fits in a little bit better with the sentence and it gets the punctuation in
(20:58):
there pretty well too.
I'm really impressed with some of these AI based transcription models lately, especially
like OpenAI's Whisper and Google's Gemini Pro 1.5 can do some really good transcripts.
That's why I'm a big fan of AI-based transcripts. And again, this is using AI to make something
(21:19):
from your content, not make the content for you. I'm not a fan of that kind of generative
AI, but I love it when you can use the AI on your content.
Keep in mind that most likely any tool that requires internet access has limits to how
much you can transcribe or has pricing based on your usage.
(21:39):
So as an alternative to those tools that I recommended, and they are great tools, I do
recommend them, I do like them, I use some of them, you could consider two fantastic
apps that you can install on your computer to generate the transcripts without the internet
or usage fees or limits and without the even possibility of your content feeding an AI somewhere.
(22:04):
Which most of these tools don't do that anyway, especially if you're paying for the tool.
But just in case you're worried about that, you can know that if you're running the AI
on your computer, your information is safe.
Both of these tools that I recommend for installing on your computer and running on your computer
use OpenAI's free and open source Whisper model for transcripts and they work really
(22:26):
well. The first is Mac Whisper Pro. The Pro version is paid but there is a free version
but I like the Pro version. It does transcripts a little bit better. It uses a bigger, large
language model which allows it to be more accurate and it's really not that expensive.
It's only around $25. I bought it. I use it myself. It's what we use for the transcripts
(22:48):
for The Audacity to Podcast, instead of some of these other services that I've noticed
kind of struggle on certain things that I say, like Podcasting 2.0 is one of those things
that other systems struggle with, or write out kind of strangely because those systems
might not know what Podcasting 2.0 is. But surprisingly, MacWhisperPro frequently properly
(23:09):
writes Podcasting 2.0 and even capitalizes the word "podcasting". How does it know
Well, because it knows enough, it's a recent enough model that it has probably seen the
phrase "podcasting 2.0" and it knows to capitalize that as a proper noun.
So I really like MacWhisperPro and things that are based on the WhisperAI model.
(23:32):
I haven't worked a whole lot with Gemini, but that is an internet-based service, whereas
MacWhisperPro runs entirely on your Mac.
On the Windows side, there's actually a completely free version.
I'm not on Windows so I can't really speak to the user experience of this, but what I've
seen it seems pretty easy to use and that is Whisper GUI.
That's for Windows, completely free.
(23:53):
I've got the link for both of these in the notes at theaudacitytopodcast.com/transcripts.
Keep this in mind.
Because these apps run everything on your computer, you'll get much faster results
on a more powerful computer.
And especially if your computer is using a more modern architecture.
For comparison, MacWhisper Pro's largest language model on my M1 MacBook Pro, so it's
(24:18):
not the latest generation of Macs, but it's still fairly new and it has that M1, Apple's
own Apple Silicon, that can transcribe a 30-minute podcast episode in only a couple of minutes.
But my maxed out Intel iMac, the last Intel iMac that Apple made with almost fully maxed
out specs, takes much longer to process the same audio with the same app and the same
(24:45):
model. And that's just because the architecture of the chip inside of the computer is different.
So some stuff you might find runs much better on something like an M1 Mac or M2 or M3 or
M whatever number than on an equivalently priced or seemingly equivalently powered Intel
(25:06):
But your results may vary. Just keep that in mind though that it may take longer
So you might have to decide is it worth the wait to get this for free?
There is speaking of free there is also a way to run whisper from the command line
But the phrase command line itself might have scared you off. It's okay. This is a safe place
(25:27):
You can come out from behind that rock
I'm not going to talk about how to do it from the command line
but I can link to some more resources if you want to go that route.
But I really recommend an app that can help you with this because I found also the app can do some nice things.
Even like with Mac Whisper Pro, one of the things I really like about it is that I can enter a saved find and replace list.
(25:48):
For example, sometimes it does transcribe The Audacity to Podcast.com with a number 2 in there
or it does other weird things every now and then.
But I can, as I've recognized a pattern of it frequently does this, I can enter those
into the list of find and replace operations and replace those with the proper URL so that
(26:08):
any time it incorrectly transcribes it, then it can go back and correct it.
Some apps will let you enter certain keywords or terms, names, proper nouns, that kind of
thing that then help the AI know whenever it hears this thing that's close to what you
entered, you probably meant the thing that you typed. And that's another thing where
(26:29):
the AI is really good at making these transcripts.
With the launch of iOS 17.4, you can also now download the transcript that Apple automatically
generates for you through your Podcast Connect account. And if you don't have access to
your podcast through the Podcast Connect account, you need to reclaim ownership. Do not resubmit
(26:50):
your podcast. I did a whole episode about that previously, so I hope that you know never ever
to do that. Just reclaim ownership of your podcast. But those transcripts will be available
for you to download in VTT format. You can download those, make some alterations if you
notice there are some problems with it, and then re-upload it in the right place. But speaking of
editing, that goes on to step four of using podcast transcripts. Edit your transcripts.
(27:15):
Regardless of whether you transcribe your podcast with AI or with a person, the transcript will most
likely need some editing. Some automatic transcription tools will smartly break lines
at logical points, like on punctuation or at the ends of sentences, but that's not mandatory for you
to do with your properly formatted, time-stamped transcript, but it can be kind of nice. It's just
(27:38):
not required. The best thing to do would be read your entire transcript to edit it for
accuracy. And I think it could even be okay to edit it for clarity. For example, if the
speaker said, "There are five ways, no, no, I mean six ways to do this," you could probably
edit that part in the transcript to simply say, "There are six ways to do this." Even
(28:03):
Even though that's not exactly what was spoken, that is the meaning of what was spoken
and you're not significantly changing the timing and you're not at all changing the
content of it.
You're just doing a tiny little bit of editing just like you might do with the audio or video
of your podcast.
Pay close attention to anything that could cause you legal trouble too.
For example, your guest might have said "I love two little puppies" but your transcript
(28:29):
might incorrectly say "I love to kill puppies". Yikes! There could be other instances too
that could be even worse than that. But this is where I notice that good modern transcription
tools like Whisper, Gemini Pro, and similar that use updated large language models or
LLMs can often do a better job because they seem good at combining what they think they
(28:53):
here, along with what makes grammatical and contextual sense, along with what they know
about what's on the internet and what content they know about.
But if you're in a hurry with your transcript, here's the quick hack for editing your transcripts.
Double check all proper nouns and URLs. And that might be the most important thing to
(29:16):
do. Because your URLs will display, and those can very easily be incorrect. And check those
proper nouns. Make sure they're spelled correctly, they are capitalized properly,
maybe there's something to with a space or without a space, that kind of stuff. For
example, one transcription tool I tried would always transcribe "podcasting 2.0" as
"podcasting" the number 2, the word "point" and then the word "oh" that's "oh"
(29:45):
and when you see that in writing it just it looks horrible. Even though that's how I'm
saying it, "Podcasting 2.0", that's not what the transcript should say. It should
be capital P in "Podcasting" and then number 2, a point or a decimal, and then a
zero. That's how it should be written. Or even, as I mentioned earlier, Mac Whisperer
(30:06):
Pro sometimes transcribes The Audacity to Podcast.com as something like the Audacity 2, number 2
there, Podcast.com or something else kind of like that, sometimes with a space, without
of space. Sometimes the systems know that certain things, like maybe Microsoft.com,
you're never going to see it transcribe Micro-space-soft.com because they just know and they understand
(30:31):
the language and somewhat understand the culture. For these things, I've made a list of the
common errors so that, like with MacWrisper Pro, it can automatically process those things.
But maybe for you, if you're not using a tool that has an automatic find and replace or
lets you enter in some of those proper nouns like podgagement so it knows to align those
(30:52):
with some of the stuff that you've written ahead of time.
Maybe you just maintain a list, a document on your computer of "this is my find and
replace, I need to always check my transcript for these kinds of things" and easily change
those so that when you're editing your transcript you can just copy, find, replace, copy, find,
replace, copy, find, replace like that and then you've fixed most of those proper nouns.
(31:16):
But I've been surprised, especially by Whisper and Gemini Pro 1.5, how often it not only
gets the spelling right, but even the capitalization, or even some made up words. Like Podgagement.
I tried this out with, right now I'm using Whisper for Podgagement on the new transcription
(31:36):
feature that when people send you a voicemail through your feedback page, it's automatically
transcribed that's using the OpenAI Whisper model. And in my trying this out, I spoke
the word "podgagement" into voicemail. And I was quite surprised it transcribed it
and capitalized it correctly. I don't know how it knew that, because "podgagement"
(32:01):
didn't exist as a word before I announced it in November 2023. So at least the internet-based
OpenAI Whisper like I use for podgagement, maybe it's using a more updated large language
model so it does know the word podgagement, or maybe it's just that good at realizing
(32:21):
what I was saying and assuming this sounds like a proper noun because it's a made up
word and therefore it should probably be capitalized. It didn't even try to capitalize the G in
there, like the only letter that should be capitalized in podgagement is the P. So that
it got that correctly was a big surprise to me. But that's the cool thing I think about
these AI-based systems. But still, don't let the coolness make you think you can completely
(32:47):
trust it. It's still best to go through and look it over, especially if you are in some
kind of industry where if you don't say something correctly, you could be in trouble or your
guest might be in trouble.
So that's step four, edit your transcripts.
Step five, put the transcripts in the right places.
(33:08):
Now that you have an accurate transcript in a good format, that would be SRT or preferably
VTT, we've had a standard place to put that in podcasting for a couple of years now and
even Apple Podcasts is now on board with this.
You only need to upload the transcript somewhere online if it's not already publicly hosted
(33:29):
for you and link to it in the podcasting 2.0 podcast colon transcript tag in your RSS feed,
which your podcast publishing or hosting tool might actually easily do for you.
And if you want more of the technical information for that, I've got a link to podcasting2.org
in the notes for this episode so you can check that out, learn more about the actual technical
(33:51):
specification if you want to, but you shouldn't have to because your publishing tool should
offer this to you and it should support Podcasting 2.0, especially now that Apple is supporting
the Podcasting 2.0 transcript tag. If your publishing tool doesn't support the transcript
tag by now, you need to switch baby. And I don't know why I just called you baby, I
(34:12):
guess because I didn't want to call you Shirley. This makes the transcript visible to a growing
list of Podcasting 2.0 apps and now even Apple Podcasts supports the Podcasting 2.0 transcript.
So it's no longer just the fringe thing that only the geeks are doing.
Even if Apple Podcasts generates the transcript for you, Apple Podcasts will use your transcript
(34:34):
if you provide it through the transcript feature in your RSS feed.
Even if you change it after you've published your episode, Apple will switch back over
to your transcript.
And I highly recommend you provide your own transcripts because then you can ensure they're
more accurate.
You could download what Apple automatically generates for you after you've published
your episode, but then you have to wait for Apple to pick it up, to download, to make
(34:59):
the transcript, and then you download it, edit it, and all of that.
That all happens after you've published the episode.
So that's why I'm suggesting you generate your transcript before you publish, then you
don't have a lot to do after the episode is published, especially if you schedule your
episode for publishing early in the morning like I do.
And this way that Apple does it, where they generate the transcript unless you've provided
(35:20):
a transcript to them.
This is so much better than how Spotify, Google, and Amazon currently do it where they generate
the transcript and that's it.
They only use theirs.
They don't use the one you provide to them.
At least not yet.
Apple is doing this the smart way.
I love that they're doing it this way.
They make one for you unless you give them one and then they use the one you give them.
(35:43):
Putting your SRT or VTT transcript in the right place also exposes it for other apps
and services to use in some cool ways.
As this becomes more popular, more podcast apps will use these transcripts to learn what
your episodes are about and potentially help expose your podcast in more relevant searches.
(36:04):
For example, even if you keyword stuff your title, author, and description tags with something
like "real estate investing" for example, if you're not literally talking about real
estate investing in your episodes, your podcast probably won't rank well for that term.
And it shouldn't if you're not actually talking about that thing.
What makes this podcast SEO different from the SEO myth that I shared previously is that
(36:29):
this podcasting 2.0 method exposes your transcript in a specific structure that makes it easy
for applications to read and process.
You don't get that when you dump the whole transcript on your website or on a download
and it's just a giant block of text or maybe some headlines, but if it's not properly
(36:49):
formatted like an SRT or VTT, then the systems can't read it and interpret it as well.
There are a lot of other cool things that apps can do, and some of them are starting
to do with things from the transcript, like allowing you to more easily clip your favorite
part of an episode, because instead of having to work with the waveform, like some old apps
(37:10):
out there that are long dead like Clamor, you could instead just say, "That thing I
just heard.
Clip that."
And it will recognize the start of the sentence.
It might even recognize the start of the value, like some of these AI tools, and I'll be playing
with some of them this year to see what kind of clips it offers me from my own content.
(37:31):
But then you might get that as a listener to be able to then share clips of the favorite
things you heard from the episodes you listened to.
But if you really want to publish a transcript in a readable format, and please do also publish
in a structured format, you can convert your SRT or VTT transcript into formatted text
(37:51):
or maybe even download it as formatted text from whatever created the transcript for you,
and then offer that through your website.
Just don't expect much to come from it.
In fact, if you are between the choice of dumping the transcript on your website as
quote "show notes" unquote or running your transcript through AI and having it generate
(38:14):
an article for you, I think actually the AI workflow would give you far better content
that would also perform better in SEO.
And remember, this is not generating content for you.
You have generated the content.
You would simply be using the AI to reformat and restructure the content and make it more
(38:34):
readable.
And that's what AI is really good at doing.
So some of these tools like Cast Magic, CapShow, Buzzsprout co-host AI, and some of these other
tools that I've mentioned that can do things with your transcript, then you can use that
for your notes on your website instead of just dumping the transcript.
(38:55):
And I think that would be more readable, it would be more memorable, more actionable,
It would be better for SEO, better for your website, even better for accessibility than
dumping the whole transcript and that whole back and forth between Jack and Jill as they
go up the hill to fetch a pail of water.
But make sure, even if you're using the AI tools, make sure you're publishing the transcript
(39:16):
in the right place.
That's step five.
And then lastly, step six, this one's optional, use your transcripts for promotion.
I'm not a fan of creating new content with AI, also called generative AI.
I think the results are cheap, unoriginal, and even in a legal grey area because most
generative AI models were trained with copyrighted content and without the consent of that content
(39:44):
owner.
That's concerning.
I know some people would say things like, "Well, I could read a whole bunch of books
and then write something based on what I learned from those books."
And yes, that is kind of true, but it would produce something completely different still.
And the thing with the AI, since it is a system, a computer, doing it, it has the ability to
(40:06):
directly copy things and simply word spin. Whereas if you are writing from your knowledge
of the books that you've read and articles and such, you might not remember things verbatim.
And besides that, it would come out in your own voice as you apply things. Like, I don't
even know how much of my knowledge that I convey is original to me versus something
(40:28):
I learned somewhere else or pieced together from other things. I'm not intentionally
piecing together other content. I'm giving content to you that's maybe even inspired
by completely irrelevant content. Not irreverent, irrelevant. But putting aside those potential
legal and ethical implications, I do love artificial intelligence as a different kind
(40:50):
of AI, assistive intelligence. So instead of creating new content for you, the AI analyzes
the content you created and helps you describe it, improve it, repurpose it, and more. And
this all starts with a transcript of your podcast. My favorite AI tool right now is
(41:10):
Magi because it includes multiple models, not just ChatGPT, and it's so much better
than ChatGPT in many ways, and it even has ChatGPT built in, among many other things
that you can do with it. And you can now, just as of a couple of weeks ago they launched
this feature, you can now easily upload your entire podcast transcript, even if it's
(41:31):
the SRT and VTT format, which I actually would recommend that you use, you can upload that
into Magi and then work with that inside of Magi. Magi doesn't have to generate the
transcript for you. Actually, right now, as of when I'm recording this, Magi can't generate
transcripts, but I would expect that would come sometime in the future. But for example,
(41:52):
you could upload your transcript to Magi, then ask it to write a promotion for the episode
in the first person tone of your guest, so that you can then send that to him or her to more
relevantly share their appearance on your podcast already pre-written in first person, so it sounds
like they are the ones saying it. That's some really cool stuff that you can do with AI
(42:16):
repurposing and reformatting your content that you give it to do that to. Cast Magic, CapShow,
and Buzzsprout's co-host AI also have features like this built in. For example, Cast Magic,
which I also use, automatically recognizes separate speakers in the audio and can also
automatically generate a whole bunch of different content from my content. This includes social
(42:42):
posts, titles, ex-Twitter threads, engaging questions, outlines, and more. And it can
do all of that after it's uploaded, after it makes the transcript, with literally one
click. And it then just starts generating all of these things. And I can even make my
own little templates in there. And I've done that to say this is the kind of thing
I want you to automatically generate for every transcript you make for me for every episode
(43:07):
that I upload.
Even if a tool can't transcribe your content for you, as is the case with Magi, you can
still give it your transcript to analyze and use it for many more purposes.
And it all starts with the transcript of your content.
Or in my case, since I write such thorough articles for my podcast episodes, I frequently
(43:29):
give Magi my published webpage URL instead of uploading a transcript. Because the article
that I already wrote, I think, is actually better than the transcript would be and might
even make the AI work a little bit better. And in case you're wondering, the entirety
of this episode and the article that accompanies it were created solely by me and even though
(43:51):
some people might call my intelligence artificial, I didn't use any AI for this episode. You
You can also use a tool like Descript, Headliner, or Opus Clip to find good excerpts for your
episodes to share as clips online, especially if you've recorded video.
And that's what I'm doing for this very episode.
I pay for Descript and I signed up for it.
(44:12):
I might sign up for Headliner and Opus Clip and try some other tools like them as well
so that I can do this kind of thing.
Record a video of me doing my podcast episodes.
I may not publish that full video, but I can use that video, upload it to these systems,
or give it to Descript, then it will transcribe my content, and then it will find little excerpts
(44:35):
that I can use and export those as videos that I can put on xTwitter, on Instagram,
on Facebook, on YouTube Shorts, or different places like that.
And all of that starts also with the transcripts.
And if possible, to upload your own transcripts to some of these tools, you might even get
better results if you do that and you upload an edited version that's more accurate before
(45:01):
you let the AI tools then do their thing.
And their thing often seems like magic sometime.
I know this was a lot of information.
I had a lot to say about transcripts.
Go to the notes at theaudacitytopodcast.com/transcripts to review this, to get the links that I mentioned
Again, these six steps for how to use podcast transcripts are step one, don't believe
(45:26):
the myths about podcast transcripts.
Optional step two, use transcripts with podcast production tools.
Step three, make properly formatted transcripts.
Step four, edit your transcripts.
Step five, put the transcripts in the right places.
And optional step six, use your transcripts for podcast promotion.
(45:47):
Get the links, the notes for everything I talked about in this episode at theaudacitytopodcast.com/transcripts.
I'd also love it if you would try out Podgagement to help you engage your audience and grow
your podcast now with transcripts.
Not in like just a "now it's got electrolytes" kind of thing, but it's a very useful thing
(46:07):
that if you want to get voicemails from your audience, as I would love to get voicemails
from you that I can use on the podcast.
I've done that in the past.
For a while I didn't do it.
I'd love to do it again.
You can do that same kind of thing with your own podcast.
Go to podgagement.com, sign up for a trial there
if you haven't tried it before,
get on the constellation plan
so that you can get the voice feedback feature
(46:31):
and get those voicemails automatically transcribed.
And you know what?
Maybe you could even use those transcripts
to run them through some AI
to do something cool for you with your podcast.
I don't know, maybe I'll build some of that
into Podgagement as well in the future.
I'd love for you to try Podgagement over at Podgagement.com.
Now that I've given you some of the guts and taught you some of the transcripting tools,
(46:54):
it's time for you to go start and grow your own podcast for passion and profit.
I'm Daniel J. Lewis from theaudacitytopodcast.com.
Thanks for listening.
[MUSIC PLAYING]
How to use podcast transcripts.
(00:09):
Thank you for joining me for the Audacity podcast.
I'm Daniel J. Lewis.
Transcripts have long been promoted as ways to make your podcast more findable and accessible,
but it's only recently that podcast transcripts have become actually useful.
I've wanted to talk about transcripts for years and I've talked about them in other
(00:29):
places and in social channels and such, but never here on The Audacity to Podcast.
So it's finally time to get to it.
And the timing of this is great because although I didn't plan the timing of these things
to line up this way, I just released a new feature for Podgagement that transcribes the
voicemail you receive from your audience through Podgagement.
(00:52):
If you want to send me a voice message, go to castfeedback.com/audacity.
You can ask me a question that I can address in a future episode, or you can just thank
me for the podcast.
If you want to just try out the system and see how it works from the listener's side,
that's castfeedback.com/audacity.
And then when you send a voice message through that system, you could also write your message
(01:14):
too if you wanted to, or you can attach a written message to your voicemail, which is
great for if you have anything that's hard to pronounce, the spelling of something is
ambiguous or you mention any URLs, it's great to be able to include that with the written message.
But when you send that, then I will receive a notification that includes the transcript
of your recorded voice.
(01:36):
So then I can very easily see what you're talking about, know how to categorize it for
my own podcast, or if I'm on the Podgagement site, I can easily filter through all of my
feedback based on what was spoken in the voicemails I received from you and other audience members.
So I'd love for you to try this over at Podgagement.com for the automatic transcriptions of voicemails
(01:59):
from your audience.
It's really cool.
It's a great way to engage with your audience.
That's over at Podgagement.com.
Please try it out.
The other thing that makes this great timing is the release of iOS 17.4 which brings podcast
transcripts to Apple podcasts and maybe even more exciting using the podcasting
(02:19):
2.0 spec. Huge thing there. I'll get more into that a little bit later in this
episode and also there's a separate episode of the future of podcasting where
Dave Jackson and I talked about the implications of all of that and I know I
addressed it briefly in my previous episode about the top features of
podcasting 2.0. Now if you want the links to these things that I mentioned then
(02:41):
then please go to the notes at theaudacitytopodcast.com/transcripts.
Little disclosure up front, several of the links on my site and things that I talk about
are affiliate links so I earn from qualifying purchases only through my links.
I am not currently sponsored by anyone so I don't earn money simply by talking about something.
(03:02):
I earn money only through some of my links, but I hope you know by now that I recommend
things I truly believe in, regardless of earnings. And in fact, some of the things I recommend
on the site don't earn me anything. So check out the notes, follow along if you'd like
a tap or swipe away or at theaudacitytopodcast.com/transcripts.
(03:23):
So how do you use podcast transcripts? Step 1. Don't believe the myths about podcast
transcripts. For many years, people have been promoting transcripts for the wrong reasons.
I've also been pitched by many transcription companies pushing the wrong approach.
While there are elements of truth to these two common myths I'm about to share, the
(03:46):
ways people often promote and understand these two ideas are where the information becomes myth.
Or could we say myth information?
Right?
Right?
Oh, come on, come on, please tell me that made you chuckle at least a little bit.
Send me a voice message through castfeedback.com/audacity to let me know what you thought of that.
(04:11):
Maybe we should just call it from now on, this is myth information, which is very hard
to say without saying it with a complete lisp.
Anyway, myth number one.
Transcripts are good for podcast SEO.
No they aren't.
People have often said that you should publish transcripts on your website for your podcast
to make your podcast findable through search engine optimization or SEO.
(04:35):
But that's not the full truth.
Yes, transcripts are better than nothing, or a mere paragraph or a list of topics through
bullet points, but transcripts result in a lot of words and actually very little content,
resulting in very low quality writing.
Which is probably even worse than an AI could create, or worse than what you might get from
(04:59):
someone who's not a native of the language. For example, here's a transcript of a two
co-host podcast that I've made up. This does not come from or it's even inspired
by anyone's podcast. So if this sounds like it's exactly your podcast, I did not listen
to it to make this fictional transcript. But it might sound something like this. And you
(05:20):
can read this in the notes if you want, but I'll read it for you.
Jack, welcome to our podcast where we talk about things to help you do stuff.
I'm Jack.
Jill, and I'm Jill.
Jack, before we get into this topic, how are you doing Jill?
Jill, the weather is sunny today so I'm great.
Jack, great!
Jill, yeah, what about you?
(05:41):
Jack, I'm fine but I don't know what the weather is but…
Jill, that's sad.
Jack, what do you mean?
Jill, I mean, you're saying that you haven't gone outside or even looked out a window in
a while.
"Yeah, but it's okay, Jill."
"So anyway…"
"Jack."
"Yes, let's jump into this week's topic."
Now note, with this fictional example, there are a lot of words.
(06:04):
But what did you actually learn from this?
What value did you get from this?
Probably nothing.
That kind of back and forth is fine in conversation, so I'm not saying don't have conversations
like this.
Sometimes even okay to be in a podcast, in appropriate places, and if done well.
(06:24):
But in writing, it becomes basically worthless.
Compare how that same "information" could be written in a way that's actually turning
that content into something valuable.
And I quote.
(06:45):
That's right.
Nothing.
That whole exchange could be edited completely out of the written content because there's
little to no value in reading that.
There might be some value in hearing that, and certainly in having that conversation,
maybe, but it's not really valuable to read that.
(07:08):
Search engines prioritize high-quality content that is readable.
But unless you're a professional speaker performing a refined monologue, transcripts
are neither readable nor high quality.
So no, transcripts are not the magic bullet to making your podcast perform well for SEO,
especially if you're relegating transcripts to a downloadable file or a webpage separate
(07:32):
from your podcast episode.
That's just not really helping you out all that well.
There is an aspect of truth, and certainly transcripts are better than nothing, but they're
not the magic bullet that everyone seems to be saying, "Yes, you should do transcripts
for SEO.
That's the reason you should do it, because it makes your podcast findable."
No, that's myth information.
(07:53):
So hard to say that without lisping overly.
Myth number two, transcripts make your podcast accessible.
No, not really.
It's true that people with hearing impairments still consume podcasts.
Thus, transcripts have been lauded as the accessibility solution all podcasts need.
But the problem is that most published transcripts are still difficult to read.
(08:17):
Like that example I shared with you.
That's not really easy to read.
Or worse, those transcripts are buried in a link or downloadable file that could actually
be even less accessible.
It's like putting up a billboard with a QR code that drivers can scan with their phones
to watch a video to learn why they shouldn't be distracted while driving.
(08:40):
It just defeats the purpose of it.
Also, good accessibility doesn't help only the impaired.
It can also help memorability or engagement for the rest of your audience.
For example, a word or URL that might be ambiguous when spoken, such as 2, 2, 2, and 2, or let's
Let's throw in a bonus one, 2, and that would be T0 vs. Two vs. The number 2 vs. T00 vs.
(09:09):
Let's throw in the Roman numeral 2, which would be two capital I's.
Which one are you actually saying?
Well, if there's a transcript that goes along with the podcast, someone could glance
at that transcript if it's properly displayed, more on that in a moment, and they could see
exactly which one is displayed.
Did you know I actually have the audacity number2podcast.com just in case someone somewhere
(09:35):
out there automatically transcribes me or writes my URL the incorrect way with a number
2 in it instead of the word T-O in it because it is a little ambiguous.
Yes, but that's a small investment of $10-$15 a year to make sure that I have that domain
so if someone gets it wrong it still takes them to the right place.
(09:56):
But that kind of ambiguity gets even worse if you've made up words.
Like podgagement is a completely made up word that I have the trademark for.
And some people might be wondering, "Wait, did he say podgagement or patgagement?"
And don't worry, yes I did register patgagement.com just in case someone mishears it or mistypes it.
(10:19):
But here, properly written show notes or an article can make the information far more
memorable and actionable than a giant transcription page.
But it's only properly formatted and properly published transcripts that make your podcast accessible.
And more on that in a moment.
But simply publishing a transcript doesn't instantly make your podcast accessible.
(10:43):
It's how you do it that makes the difference.
Those are the two myths I hear and see most often.
Now as you've noticed, there is an element of truth to them, but most of the time, the
way that people are promoting and understanding these things makes them, I just gotta say
it again, makes them myth information.
Oh, I'm way too entertained by that.
(11:05):
Step 2, and this one's optional, use transcripts with podcast production tools.
You can use the power of transcripts immediately after you record your episodes, even if you
never publish those transcripts.
While you could pay for someone to transcribe your podcast for you, AI tools have gotten
really good at making fairly accurate transcripts. And some of them can even listen while you
(11:30):
are recording and give you the transcript immediately after you finish recording.
Yes, AI has become the ubiquitous tool here and really for all kinds of content creators.
I've joked before in multiple places the line from It's a Wonderful Life, a classic Christmas
movie you gotta watch. I watch it every year. I've started crying more frequently now when
I watch it. But there's the classic line in there and here's my version of it, modernized.
(11:54):
"Look daddy, teacher says every time a bell rings, an AI-powered app launches on AppSumo."
Isn't that the truth?
While I'm not a fan of some of these AI apps that are coming out and how some people are
leaning too heavily on AI, I am a big fan of using AI on content you made instead of
(12:15):
using it for making your content for you.
The way AI-based tools work with your audio or video podcast is by first transcribing
your content. That's where the transcripts come into play here. Then, you can use the
AI to do impressive and time-saving things with that transcript. Like summarizing your
(12:35):
episode, suggesting episode titles, pointing out the best parts of your episode, or even
helping you better edit your recording to find the stuff that wasn't of much value
or find the stuff that you want to pull out and use in particular places.
There are a bunch of tools out there, and like I can easily joke, it seems like a new
AI powered tool is launching every day, throughout the day, on some of these startup sites.
(12:58):
But here are my current favorite transcription based tools to help when your episode is still
in this early production stage, even right after you've recorded.
It works great for audio and video.
And in fact, I will be using Descript to not edit the audio of this episode, but to pull
out some video clips of this episode.
(13:19):
I'll talk more about that idea in a moment.
You could also use Adobe Podcast for editing your podcast episode.
That works only on audio currently.
Or Adobe Premiere Pro works on video and you can actually even edit audio inside of Adobe
Premiere Pro because it has a transcription based editing tool built into it.
And Adobe Audition...
(13:41):
Someday maybe sort of?
Audition seems to follow behind Premiere Pro by maybe about a year on some of the features.
So I expect to see some of this cool stuff that Adobe is doing like with Adobe Podcast
and Premiere Pro.
I expect to see that coming into Audition and it's already in there in some places actually.
And the other thing is my favorite podcast editing app, Hindenburg.
(14:04):
I get paid nothing for recommending them.
They don't have an affiliate program.
I want them to have one.
Still, it's my top recommendation for an audio editing app for podcasting because it
has so many great things and especially with version 2, it now has transcription based
editing and you can export that transcript and use it for all kinds of cool things.
So each of these tools that I mentioned, and this could be an ever-growing list, I'm not
(14:28):
going to try and keep this list up to date because AI tools come and go, but these are
the most popular, the ones I expect to stay around. They're the most professional tools
as well. Each of these tools can make editing your recording as easy as editing a text document.
So that's why you might frequently see me recommend, especially Descript, recommend
(14:49):
that as a solution for someone just getting into podcasting who knows nothing about audio
editing. But besides just editing the audio or the video with the transcript, these apps
can also export the transcript for your finished production that you can then use in some of
the other steps I'm going to share in this episode.
(15:10):
Even if you don't use any transcription based tools, you can still use a transcript
of your recording to help you find places to edit, excerpts for sharing, and reminders
of your content, like when you're writing your notes or writing those posts that are
going to promote your episode, or maybe sending something to your guest, you can look back
over the transcript to remember, "Oh yeah, that's what we talked about," or "Oh yeah,
(15:32):
I need to make sure to include that URL in the notes for this episode because that came
to me just spontaneously while recording this episode.
I need to make sure that's included in the notes."
It can really help you do some of those things much faster simply by having the transcript,
even if you don't publish it.
But I still recommend that you do publish your transcripts.
(15:53):
More on that in just a moment, but if you want the links to these tools that I mentioned,
get them in the notes for this episode a tap or swipe away or at theaudacitytopodcast.com/transcripts.
Step 3 to using podcast transcripts.
Make properly formatted podcast transcripts.
In order to be actually useful, a transcript needs to have more data with it.
(16:16):
At a minimum, transcripts need timestamps for when lines or even words were spoken.
That's what enables closed captioning to work.
Closed captioning is not just a giant block of text in a document somewhere.
It is timestamped text.
So the system that's playing the video knows exactly when to display what lines of text.
(16:38):
That's the same way it works in a properly formatted podcast transcript.
Transcripts can also include the speaker's names, formatting, and more. The most common
transcript formats are VTT and SRT. But VTT, and specifically WebVTT, is the superior format
(16:58):
and what I recommend now because of some cooler things that you can do with it. In its vanilla
state VTT and SRT will actually look similar, but WebVTT allows you to do a whole lot more
on top of the normal timestamped transcript. If you look in the notes over at theaudacitytopodcast.com/transcripts
(17:20):
you'll get to see a little example of what one of these VTT or SRT transcript files will
look like. Where you have a time code range, so it's two timestamps, start and end, one
line of text underneath it, and then another timestamp range, and then another line of
text. Sometimes you might see multiple lines. You might even see some formatting in there if it's
(17:43):
WebVTT. But that's what it is. It's the transcript broken down into lines, sometimes maybe even
individual words, with timestamps for that information. And then WebVTT can add a whole
new layer on top of that with additional features and formatting and a lot of other really cool
stuff. So that's why I think VTT is the superior format and what I'll be recommending and switching
(18:08):
to from now on. But don't let the format of these transcripts scare you off because you can get
these properly formatted transcripts from the transcription-based editing tools I shared
previously, or you can generate them through AI or third-party help. And remember the most
important thing here is to get the transcripts in the right format with those timestamps in that
(18:32):
actual SRT or VTT format. A big block of text in a PDF or text document is not going to be as useful
as the structured data like you get with SRT or VTT and that structure is necessary for all of the
cool magic that can happen in podcast apps now thanks to Podcasting 2.0. Easy access to artificial
(18:56):
intelligence or AI tools has made creating transcripts a whole lot faster, easier, and
cheaper. The editing tools I mentioned all use AI to generate their transcripts. Or you could use
other tools and services too, like Cast Magic, CapShow, Affonic, Buzzsprout's co-host AI add-on,
Transistor now has an AI-based transcript tool built in, Otter, Rev, and other tools here and
(19:21):
there that can help you generate those transcripts and some of these tools can do more things with
those transcripts. This list also is not going to be up to date forever because a lot of other
tools come and go with these cool transcript based features that have AI built in. But these are some
of the best ones I think right now. I like the more modern AI based transcription tools because
(19:45):
they understand language better. They're called large language models, or LLMs. And
because of that, they're able to make transcripts that actually read well and are usually quite
accurate, even with pretty good punctuation too. Compare that to transcripts from only
a few years ago that transcribed what they thought they heard, even if it didn't make
(20:10):
You can look up on YouTube stuff like "caption fail" or "transcription fail" and I used
to do this and make fun of this myself with my comedy podcast where we would use Google
Voice which would transcribe the voicemail, but the transcripts were just what it thought
it heard even if it didn't make sense.
And so we could sometimes read those transcripts instead of playing the voicemail and reading
(20:34):
it just did not make sense at all.
It would insert a random word here and there just because it thought that's the word it heard.
But with these AI systems that know how to work with language, they're much better at
understanding this is probably actually what you meant and it's pretty close to the sound
and this fits in a little bit better with the sentence and it gets the punctuation in
(20:58):
there pretty well too.
I'm really impressed with some of these AI based transcription models lately, especially
like OpenAI's Whisper and Google's Gemini Pro 1.5 can do some really good transcripts.
That's why I'm a big fan of AI-based transcripts. And again, this is using AI to make something
(21:19):
from your content, not make the content for you. I'm not a fan of that kind of generative
AI, but I love it when you can use the AI on your content.
Keep in mind that most likely any tool that requires internet access has limits to how
much you can transcribe or has pricing based on your usage.
(21:39):
So as an alternative to those tools that I recommended, and they are great tools, I do
recommend them, I do like them, I use some of them, you could consider two fantastic
apps that you can install on your computer to generate the transcripts without the internet
or usage fees or limits and without the even possibility of your content feeding an AI somewhere.
(22:04):
Which most of these tools don't do that anyway, especially if you're paying for the tool.
But just in case you're worried about that, you can know that if you're running the AI
on your computer, your information is safe.
Both of these tools that I recommend for installing on your computer and running on your computer
use OpenAI's free and open source Whisper model for transcripts and they work really
(22:26):
well. The first is Mac Whisper Pro. The Pro version is paid but there is a free version
but I like the Pro version. It does transcripts a little bit better. It uses a bigger, large
language model which allows it to be more accurate and it's really not that expensive.
It's only around $25. I bought it. I use it myself. It's what we use for the transcripts
(22:48):
for The Audacity to Podcast, instead of some of these other services that I've noticed
kind of struggle on certain things that I say, like Podcasting 2.0 is one of those things
that other systems struggle with, or write out kind of strangely because those systems
might not know what Podcasting 2.0 is. But surprisingly, MacWhisperPro frequently properly
(23:09):
writes Podcasting 2.0 and even capitalizes the word "podcasting". How does it know
Well, because it knows enough, it's a recent enough model that it has probably seen the
phrase "podcasting 2.0" and it knows to capitalize that as a proper noun.
So I really like MacWhisperPro and things that are based on the WhisperAI model.
(23:32):
I haven't worked a whole lot with Gemini, but that is an internet-based service, whereas
MacWhisperPro runs entirely on your Mac.
On the Windows side, there's actually a completely free version.
I'm not on Windows so I can't really speak to the user experience of this, but what I've
seen it seems pretty easy to use and that is Whisper GUI.
That's for Windows, completely free.
(23:53):
I've got the link for both of these in the notes at theaudacitytopodcast.com/transcripts.
Keep this in mind.
Because these apps run everything on your computer, you'll get much faster results
on a more powerful computer.
And especially if your computer is using a more modern architecture.
For comparison, MacWhisper Pro's largest language model on my M1 MacBook Pro, so it's
(24:18):
not the latest generation of Macs, but it's still fairly new and it has that M1, Apple's
own Apple Silicon, that can transcribe a 30-minute podcast episode in only a couple of minutes.
But my maxed out Intel iMac, the last Intel iMac that Apple made with almost fully maxed
out specs, takes much longer to process the same audio with the same app and the same
(24:45):
model. And that's just because the architecture of the chip inside of the computer is different.
So some stuff you might find runs much better on something like an M1 Mac or M2 or M3 or
M whatever number than on an equivalently priced or seemingly equivalently powered Intel
(25:06):
But your results may vary. Just keep that in mind though that it may take longer
So you might have to decide is it worth the wait to get this for free?
There is speaking of free there is also a way to run whisper from the command line
But the phrase command line itself might have scared you off. It's okay. This is a safe place
(25:27):
You can come out from behind that rock
I'm not going to talk about how to do it from the command line
but I can link to some more resources if you want to go that route.
But I really recommend an app that can help you with this because I found also the app can do some nice things.
Even like with Mac Whisper Pro, one of the things I really like about it is that I can enter a saved find and replace list.
(25:48):
For example, sometimes it does transcribe The Audacity to Podcast.com with a number 2 in there
or it does other weird things every now and then.
But I can, as I've recognized a pattern of it frequently does this, I can enter those
into the list of find and replace operations and replace those with the proper URL so that
(26:08):
any time it incorrectly transcribes it, then it can go back and correct it.
Some apps will let you enter certain keywords or terms, names, proper nouns, that kind of
thing that then help the AI know whenever it hears this thing that's close to what you
entered, you probably meant the thing that you typed. And that's another thing where
(26:29):
the AI is really good at making these transcripts.
With the launch of iOS 17.4, you can also now download the transcript that Apple automatically
generates for you through your Podcast Connect account. And if you don't have access to
your podcast through the Podcast Connect account, you need to reclaim ownership. Do not resubmit
(26:50):
your podcast. I did a whole episode about that previously, so I hope that you know never ever
to do that. Just reclaim ownership of your podcast. But those transcripts will be available
for you to download in VTT format. You can download those, make some alterations if you
notice there are some problems with it, and then re-upload it in the right place. But speaking of
editing, that goes on to step four of using podcast transcripts. Edit your transcripts.
(27:15):
Regardless of whether you transcribe your podcast with AI or with a person, the transcript will most
likely need some editing. Some automatic transcription tools will smartly break lines
at logical points, like on punctuation or at the ends of sentences, but that's not mandatory for you
to do with your properly formatted, time-stamped transcript, but it can be kind of nice. It's just
(27:38):
not required. The best thing to do would be read your entire transcript to edit it for
accuracy. And I think it could even be okay to edit it for clarity. For example, if the
speaker said, "There are five ways, no, no, I mean six ways to do this," you could probably
edit that part in the transcript to simply say, "There are six ways to do this." Even
(28:03):
Even though that's not exactly what was spoken, that is the meaning of what was spoken
and you're not significantly changing the timing and you're not at all changing the
content of it.
You're just doing a tiny little bit of editing just like you might do with the audio or video
of your podcast.
Pay close attention to anything that could cause you legal trouble too.
For example, your guest might have said "I love two little puppies" but your transcript
(28:29):
might incorrectly say "I love to kill puppies". Yikes! There could be other instances too
that could be even worse than that. But this is where I notice that good modern transcription
tools like Whisper, Gemini Pro, and similar that use updated large language models or
LLMs can often do a better job because they seem good at combining what they think they
(28:53):
here, along with what makes grammatical and contextual sense, along with what they know
about what's on the internet and what content they know about.
But if you're in a hurry with your transcript, here's the quick hack for editing your transcripts.
Double check all proper nouns and URLs. And that might be the most important thing to
(29:16):
do. Because your URLs will display, and those can very easily be incorrect. And check those
proper nouns. Make sure they're spelled correctly, they are capitalized properly,
maybe there's something to with a space or without a space, that kind of stuff. For
example, one transcription tool I tried would always transcribe "podcasting 2.0" as
"podcasting" the number 2, the word "point" and then the word "oh" that's "oh"
(29:45):
and when you see that in writing it just it looks horrible. Even though that's how I'm
saying it, "Podcasting 2.0", that's not what the transcript should say. It should
be capital P in "Podcasting" and then number 2, a point or a decimal, and then a
zero. That's how it should be written. Or even, as I mentioned earlier, Mac Whisperer
(30:06):
Pro sometimes transcribes The Audacity to Podcast.com as something like the Audacity 2, number 2
there, Podcast.com or something else kind of like that, sometimes with a space, without
of space. Sometimes the systems know that certain things, like maybe Microsoft.com,
you're never going to see it transcribe Micro-space-soft.com because they just know and they understand
(30:31):
the language and somewhat understand the culture. For these things, I've made a list of the
common errors so that, like with MacWrisper Pro, it can automatically process those things.
But maybe for you, if you're not using a tool that has an automatic find and replace or
lets you enter in some of those proper nouns like podgagement so it knows to align those
(30:52):
with some of the stuff that you've written ahead of time.
Maybe you just maintain a list, a document on your computer of "this is my find and
replace, I need to always check my transcript for these kinds of things" and easily change
those so that when you're editing your transcript you can just copy, find, replace, copy, find,
replace, copy, find, replace like that and then you've fixed most of those proper nouns.
(31:16):
But I've been surprised, especially by Whisper and Gemini Pro 1.5, how often it not only
gets the spelling right, but even the capitalization, or even some made up words. Like Podgagement.
I tried this out with, right now I'm using Whisper for Podgagement on the new transcription
(31:36):
feature that when people send you a voicemail through your feedback page, it's automatically
transcribed that's using the OpenAI Whisper model. And in my trying this out, I spoke
the word "podgagement" into voicemail. And I was quite surprised it transcribed it
and capitalized it correctly. I don't know how it knew that, because "podgagement"
(32:01):
didn't exist as a word before I announced it in November 2023. So at least the internet-based
OpenAI Whisper like I use for podgagement, maybe it's using a more updated large language
model so it does know the word podgagement, or maybe it's just that good at realizing
(32:21):
what I was saying and assuming this sounds like a proper noun because it's a made up
word and therefore it should probably be capitalized. It didn't even try to capitalize the G in
there, like the only letter that should be capitalized in podgagement is the P. So that
it got that correctly was a big surprise to me. But that's the cool thing I think about
these AI-based systems. But still, don't let the coolness make you think you can completely
(32:47):
trust it. It's still best to go through and look it over, especially if you are in some
kind of industry where if you don't say something correctly, you could be in trouble or your
guest might be in trouble.
So that's step four, edit your transcripts.
Step five, put the transcripts in the right places.
(33:08):
Now that you have an accurate transcript in a good format, that would be SRT or preferably
VTT, we've had a standard place to put that in podcasting for a couple of years now and
even Apple Podcasts is now on board with this.
You only need to upload the transcript somewhere online if it's not already publicly hosted
(33:29):
for you and link to it in the podcasting 2.0 podcast colon transcript tag in your RSS feed,
which your podcast publishing or hosting tool might actually easily do for you.
And if you want more of the technical information for that, I've got a link to podcasting2.org
in the notes for this episode so you can check that out, learn more about the actual technical
(33:51):
specification if you want to, but you shouldn't have to because your publishing tool should
offer this to you and it should support Podcasting 2.0, especially now that Apple is supporting
the Podcasting 2.0 transcript tag. If your publishing tool doesn't support the transcript
tag by now, you need to switch baby. And I don't know why I just called you baby, I
(34:12):
guess because I didn't want to call you Shirley. This makes the transcript visible to a growing
list of Podcasting 2.0 apps and now even Apple Podcasts supports the Podcasting 2.0 transcript.
So it's no longer just the fringe thing that only the geeks are doing.
Even if Apple Podcasts generates the transcript for you, Apple Podcasts will use your transcript
(34:34):
if you provide it through the transcript feature in your RSS feed.
Even if you change it after you've published your episode, Apple will switch back over
to your transcript.
And I highly recommend you provide your own transcripts because then you can ensure they're
more accurate.
You could download what Apple automatically generates for you after you've published
your episode, but then you have to wait for Apple to pick it up, to download, to make
(34:59):
the transcript, and then you download it, edit it, and all of that.
That all happens after you've published the episode.
So that's why I'm suggesting you generate your transcript before you publish, then you
don't have a lot to do after the episode is published, especially if you schedule your
episode for publishing early in the morning like I do.
And this way that Apple does it, where they generate the transcript unless you've provided
(35:20):
a transcript to them.
This is so much better than how Spotify, Google, and Amazon currently do it where they generate
the transcript and that's it.
They only use theirs.
They don't use the one you provide to them.
At least not yet.
Apple is doing this the smart way.
I love that they're doing it this way.
They make one for you unless you give them one and then they use the one you give them.
(35:43):
Putting your SRT or VTT transcript in the right place also exposes it for other apps
and services to use in some cool ways.
As this becomes more popular, more podcast apps will use these transcripts to learn what
your episodes are about and potentially help expose your podcast in more relevant searches.
(36:04):
For example, even if you keyword stuff your title, author, and description tags with something
like "real estate investing" for example, if you're not literally talking about real
estate investing in your episodes, your podcast probably won't rank well for that term.
And it shouldn't if you're not actually talking about that thing.
What makes this podcast SEO different from the SEO myth that I shared previously is that
(36:29):
this podcasting 2.0 method exposes your transcript in a specific structure that makes it easy
for applications to read and process.
You don't get that when you dump the whole transcript on your website or on a download
and it's just a giant block of text or maybe some headlines, but if it's not properly
(36:49):
formatted like an SRT or VTT, then the systems can't read it and interpret it as well.
There are a lot of other cool things that apps can do, and some of them are starting
to do with things from the transcript, like allowing you to more easily clip your favorite
part of an episode, because instead of having to work with the waveform, like some old apps
(37:10):
out there that are long dead like Clamor, you could instead just say, "That thing I
just heard.
Clip that."
And it will recognize the start of the sentence.
It might even recognize the start of the value, like some of these AI tools, and I'll be playing
with some of them this year to see what kind of clips it offers me from my own content.
(37:31):
But then you might get that as a listener to be able to then share clips of the favorite
things you heard from the episodes you listened to.
But if you really want to publish a transcript in a readable format, and please do also publish
in a structured format, you can convert your SRT or VTT transcript into formatted text
(37:51):
or maybe even download it as formatted text from whatever created the transcript for you,
and then offer that through your website.
Just don't expect much to come from it.
In fact, if you are between the choice of dumping the transcript on your website as
quote "show notes" unquote or running your transcript through AI and having it generate
(38:14):
an article for you, I think actually the AI workflow would give you far better content
that would also perform better in SEO.
And remember, this is not generating content for you.
You have generated the content.
You would simply be using the AI to reformat and restructure the content and make it more
(38:34):
readable.
And that's what AI is really good at doing.
So some of these tools like Cast Magic, CapShow, Buzzsprout co-host AI, and some of these other
tools that I've mentioned that can do things with your transcript, then you can use that
for your notes on your website instead of just dumping the transcript.
(38:55):
And I think that would be more readable, it would be more memorable, more actionable,
It would be better for SEO, better for your website, even better for accessibility than
dumping the whole transcript and that whole back and forth between Jack and Jill as they
go up the hill to fetch a pail of water.
But make sure, even if you're using the AI tools, make sure you're publishing the transcript
(39:16):
in the right place.
That's step five.
And then lastly, step six, this one's optional, use your transcripts for promotion.
I'm not a fan of creating new content with AI, also called generative AI.
I think the results are cheap, unoriginal, and even in a legal grey area because most
generative AI models were trained with copyrighted content and without the consent of that content
(39:44):
owner.
That's concerning.
I know some people would say things like, "Well, I could read a whole bunch of books
and then write something based on what I learned from those books."
And yes, that is kind of true, but it would produce something completely different still.
And the thing with the AI, since it is a system, a computer, doing it, it has the ability to
(40:06):
directly copy things and simply word spin. Whereas if you are writing from your knowledge
of the books that you've read and articles and such, you might not remember things verbatim.
And besides that, it would come out in your own voice as you apply things. Like, I don't
even know how much of my knowledge that I convey is original to me versus something
(40:28):
I learned somewhere else or pieced together from other things. I'm not intentionally
piecing together other content. I'm giving content to you that's maybe even inspired
by completely irrelevant content. Not irreverent, irrelevant. But putting aside those potential
legal and ethical implications, I do love artificial intelligence as a different kind
(40:50):
of AI, assistive intelligence. So instead of creating new content for you, the AI analyzes
the content you created and helps you describe it, improve it, repurpose it, and more. And
this all starts with a transcript of your podcast. My favorite AI tool right now is
(41:10):
Magi because it includes multiple models, not just ChatGPT, and it's so much better
than ChatGPT in many ways, and it even has ChatGPT built in, among many other things
that you can do with it. And you can now, just as of a couple of weeks ago they launched
this feature, you can now easily upload your entire podcast transcript, even if it's
(41:31):
the SRT and VTT format, which I actually would recommend that you use, you can upload that
into Magi and then work with that inside of Magi. Magi doesn't have to generate the
transcript for you. Actually, right now, as of when I'm recording this, Magi can't generate
transcripts, but I would expect that would come sometime in the future. But for example,
(41:52):
you could upload your transcript to Magi, then ask it to write a promotion for the episode
in the first person tone of your guest, so that you can then send that to him or her to more
relevantly share their appearance on your podcast already pre-written in first person, so it sounds
like they are the ones saying it. That's some really cool stuff that you can do with AI
(42:16):
repurposing and reformatting your content that you give it to do that to. Cast Magic, CapShow,
and Buzzsprout's co-host AI also have features like this built in. For example, Cast Magic,
which I also use, automatically recognizes separate speakers in the audio and can also
automatically generate a whole bunch of different content from my content. This includes social
(42:42):
posts, titles, ex-Twitter threads, engaging questions, outlines, and more. And it can
do all of that after it's uploaded, after it makes the transcript, with literally one
click. And it then just starts generating all of these things. And I can even make my
own little templates in there. And I've done that to say this is the kind of thing
I want you to automatically generate for every transcript you make for me for every episode
(43:07):
that I upload.
Even if a tool can't transcribe your content for you, as is the case with Magi, you can
still give it your transcript to analyze and use it for many more purposes.
And it all starts with the transcript of your content.
Or in my case, since I write such thorough articles for my podcast episodes, I frequently
(43:29):
give Magi my published webpage URL instead of uploading a transcript. Because the article
that I already wrote, I think, is actually better than the transcript would be and might
even make the AI work a little bit better. And in case you're wondering, the entirety
of this episode and the article that accompanies it were created solely by me and even though
(43:51):
some people might call my intelligence artificial, I didn't use any AI for this episode. You
You can also use a tool like Descript, Headliner, or Opus Clip to find good excerpts for your
episodes to share as clips online, especially if you've recorded video.
And that's what I'm doing for this very episode.
I pay for Descript and I signed up for it.
(44:12):
I might sign up for Headliner and Opus Clip and try some other tools like them as well
so that I can do this kind of thing.
Record a video of me doing my podcast episodes.
I may not publish that full video, but I can use that video, upload it to these systems,
or give it to Descript, then it will transcribe my content, and then it will find little excerpts
(44:35):
that I can use and export those as videos that I can put on xTwitter, on Instagram,
on Facebook, on YouTube Shorts, or different places like that.
And all of that starts also with the transcripts.
And if possible, to upload your own transcripts to some of these tools, you might even get
better results if you do that and you upload an edited version that's more accurate before
(45:01):
you let the AI tools then do their thing.
And their thing often seems like magic sometime.
I know this was a lot of information.
I had a lot to say about transcripts.
Go to the notes at theaudacitytopodcast.com/transcripts to review this, to get the links that I mentioned
Again, these six steps for how to use podcast transcripts are step one, don't believe
(45:26):
the myths about podcast transcripts.
Optional step two, use transcripts with podcast production tools.
Step three, make properly formatted transcripts.
Step four, edit your transcripts.
Step five, put the transcripts in the right places.
And optional step six, use your transcripts for podcast promotion.
(45:47):
Get the links, the notes for everything I talked about in this episode at theaudacitytopodcast.com/transcripts.
I'd also love it if you would try out Podgagement to help you engage your audience and grow
your podcast now with transcripts.
Not in like just a "now it's got electrolytes" kind of thing, but it's a very useful thing
(46:07):
that if you want to get voicemails from your audience, as I would love to get voicemails
from you that I can use on the podcast.
I've done that in the past.
For a while I didn't do it.
I'd love to do it again.
You can do that same kind of thing with your own podcast.
Go to podgagement.com, sign up for a trial there
if you haven't tried it before,
get on the constellation plan
so that you can get the voice feedback feature
(46:31):
and get those voicemails automatically transcribed.
And you know what?
Maybe you could even use those transcripts
to run them through some AI
to do something cool for you with your podcast.
I don't know, maybe I'll build some of that
into Podgagement as well in the future.
I'd love for you to try Podgagement over at Podgagement.com.
Now that I've given you some of the guts and taught you some of the transcripting tools,
(46:54):
it's time for you to go start and grow your own podcast for passion and profit.
I'm Daniel J. Lewis from theaudacitytopodcast.com.
Thanks for listening.
[MUSIC PLAYING]
Popular Podcasts
Bookmarked by Reese's Book Club
Welcome to Bookmarked by Reese’s Book Club — the podcast where great stories, bold women, and irresistible conversations collide! Hosted by award-winning journalist Danielle Robay, each week new episodes balance thoughtful literary insight with the fervor of buzzy book trends, pop culture and more. Bookmarked brings together celebrities, tastemakers, influencers and authors from Reese's Book Club and beyond to share stories that transcend the page. Pull up a chair. You’re not just listening — you’re part of the conversation.
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com