If you are spending the time and energy producing podcast episodes, it’s important to make your content as discoverable and accessible as possible. 

There is a huge disconnect between the amount of information available in podcasts and how accessible it is to users. The availability of podcasts does not amount to the low likelihood of us finding those podcasts and it is mostly due to transcriptions.

A podcast transcription is a document that lays out, word-for-word what was spoken during your podcast episode. Some people decide to type these themselves, and others utilize automated softwares that will do the transcriptions for them. Adding Chapter Points to these transcriptions can further advance and improve your content for your users.

As of March 2022, only 1% of today’s podcasts offer transcripts- which means 99% of podcasts aren’t utilizing the incredible opportunities podcast transcriptions bring to the table. Podcast transcriptions not only make your content more accessible to everyone but also maximize your SEO reach by creating a massive amount of content. 

{{cta}}

Generating a podcast transcript

There are a couple ways to create a podcast transcription for your audience. Writing your own transcription is pretty simple, you just have to listen to your episode and write down what you and your guests say; however, this method costs an incredible amount of time. A 20-minute episode could easily take up to three hours from pausing, rewinding, and writing everything down. 

Many decide not to opt for this option, and instead decide to use a software that specializes in word-to-text transcription services. OneAI’s Language Studio is a great example that not only specializes in speech recognition, but also summaries, topics, headlines, and much more. Today’s tutorial will focus on how to create a podcast Transcription and Chapter points with the Studio.

The Language Studio

First, let’s jump to OneAI’s Language Studio. 

The studio is in the picture below. The tool input is where you can enter your audio, alongside it is the pipeline box which includes all our Language Skills.

Underneath the pipeline box, you can see the Language Skills, which are different ways we can apply the AI to the content provided. There are different skills we can apply, such as ‘Summarize,’ ‘Proofread,’ ‘Names’- today we will be clicking on ‘Transcribe Audio.’

The purpose of this Language Skill is to create a transcript for the audio provided.

This OneAI podcast will serve as our example today: 

Let's head over to OneAI’s Language Studio and upload our audio file. 

In this tutorial we are going to get a word-for-word transcript for our podcast and split it by chapters. You can use chapters as is just placing the transcript of your podcast on the website to make it easier to browse the content. Or use them to create multiple articles on different topics raised during the podcast. Another option is to pick a chapter to share it on Twitter or facebook. 

First, to obtain the transcript, we can go ahead and double-click the box that says “Transcribe Audio” (however, it should be applied automatically). We recommend adding the skill “Proofread”- to prevent any grammatical errors and reduce filler words- so go ahead and double-click that box. 

Next, drag the “Split by Topic” box to our pipeline to get chapters. Each chapter will be named automatically using NLP skills according to different topics discussed. 

Now click the box on the top right corner titled ‘Run Pipeline’ and let’s see the results: 

Here are the Language Skills that were detected on top of those we already selected:

We are going to stop with “Transcribe Audio” and “Split by Topic” in our example. Time to copy your code and proceed to the code editor.

In the code editor:

Step 1: Pick one of the generated codes you need to your code editor (click here to generate your own API key):

Step 2: Run “pip install oneai” for Python SDK or “npm install oneai” for Node.js SDK to get the library. Make sure to import all the required packages:

```

import oneai

import base64

oneai.api_key = "[YOUR ONEAI API KEY]"

```

In our example, we used Python SDK code. Here is what we got when we run all the skills together: transcribe, proofread and split by topic:

```

with open("AudioFile.mp3", "rb") as f:

   pipeline = oneai.Pipeline(

       steps = [

           oneai.skills.Transcribe(),

           oneai.skills.Proofread(),

           oneai.skills.SplitByTopic()

         ]

       )

Output = pipeline.run(f)

```

Here are the outputs:

Transcription with proofread:

```

print(Output.transcription.proofread.text)

```

```

[

0:00:00.470000 speaker 1: Canva is a visual communications platform that gives workplaces of all sizes a unique, flexible and collaborative way to make work feel like play.,

0:00:09.170000 speaker 1: With Canva, you get to have fun and kick off every project fully inspired.,

0:00:13.550000 speaker 1: Choose from thousands of free, professional templates to help bring your best work to life.,

0:00:18.110000 speaker 1: Faster.,

0:00:19.010000 speaker 1: Whether it's a product pitch, conference keynote, or group assignment, you can customize any template to suit any project.,

0:00:25.490000 speaker 1: Plus, customize your slides with thousands of free images, videos, graphics, and more from Canva's rich content library, prepping for a presentation, Canva lets you access your slides on any browser or mobile device design, edit or present on the go and collaborate with ease from anywhere.,

0:00:42.410000 speaker 1: Invite anyone to design with you and share feedback in real time from any device.,

0:00:47.090000 speaker 1: No design skills or installations are required to use Canva.,

0:00:50.690000 speaker 1: Just you, your genius and your desire to create.,

0:00:53.990000 speaker 1: Enjoy presentations for free at Canva dot com.,

0:00:58.130000 speaker 2: Reaganomics radio is sponsored by discover.,

0:01:00.470000 speaker 2: Earning cash back on your purchases is great.,

0:01:03.530000 speaker 2: But discover will do you one better.,

0:01:05.870000 speaker 2: At the end of your first year, discover will actually match all the cashback you burned on your credit card.,

0:01:11.330000 speaker 2: That's right.,

0:01:12.350000 speaker 2: Discover matches all the cash back you earn on your credit card at the end of your first year.,

0:01:17.810000 speaker 2: This is especially amazing because is accepted at 99% of places in the U.S. that take credit cards.,

0:01:24.950000 speaker 2: Learn more at discover dot com slash yes. 2021 Nelsen report limitations apply.,

0:01:39.290000 speaker 2: Several years ago, we put out a two part series, episode number two 11 and two 12 called the economics of sleep.,

0:01:45.950000 speaker 2: We looked at the relationship between sleep and health, both physical and cognitive.,

0:01:51.170000 speaker 3: You simply.,

0:01:51.890000 speaker 4: Can not think as fast and solve a problem as quickly when you're sleep deprived is when you're not sleep deprived.,

0:01:58.130000 speaker 2: And the relationship between sleep and income.,

0:02:01.310000 speaker 1: Generally, people who have more opportunities, more control over their lives, are also better sleepers.,

0:02:08.870000 speaker 2: None of these claims were particularly surprising, or at least they shouldn't be.,

0:02:13.130000 speaker 2: Anything the human body requires for one third of its operating hours must be pretty important.,

0:02:18.830000 speaker 2: What did?,

0:02:19.250000 speaker 4: Surprise.,

0:02:19.430000 speaker 2: Us was how little good, clean, real world data there was on sleeping.,

0:02:25.610000 speaker 1: I realized that we have data on the social experiences of individuals from childhood to middle age or older ages, and we are looking at what social factors explain health.,

0:02:38.990000 speaker 1: And they had an enormous amount of data on these individuals on two thirds of their lives.,

0:02:45.950000 speaker 1: The waking hours, but they didn't have anything on what's going on during that remaining third at night.,

0:02:51.590000 speaker 2: Why don't we have better sleep data?,

0:02:54.410000 speaker 1: It's quite difficult to get accurate information about people's routine sleep behavior.,

0:02:59.990000 speaker 2: What would it take to get that kind of information?,

0:03:03.110000 speaker 2: What we really need.,

0:03:04.670000 speaker 3: Is something like an experiment for sleep.,

0:03:08.090000 speaker 2: We did learn about one sleep experiment that was just getting underway in Chennai, India, a city of 10 million people.,

0:03:15.230000 speaker 2: The researchers wanted to explore the relationship between sleep and labor productivity.,

0:03:20.450000 speaker 2: Here's The Economist Heather Schofield.,

0:03:22.910000 speaker 1: If people are tired enough to be sleeping in the middle of the street and the 100deg heat with trucks going by, it's pretty hard for them to be as productive as possible in the labor force.,

0:03:33.350000 speaker 1: So if we can help improve their sleep, our hypothesis is that it will very much improve their ability to work longer and work harder and work better.,

0:03:42.470000 speaker 2: Today, on Reaganomics radio, we circle back to that experiment and tell you the findings.,

0:03:48.530000 speaker 4: We were shocked initially to find these results.,

0:03:51.170000 speaker 2: Also, we explore a particular mode of sleep to see whether it can help.,

0:03:56.990000 speaker 5: There are some departments that have implemented nothing policies.,

0:04:00.350000 speaker 2: Napping policies, wait a minute.,

0:04:02.330000 speaker 2: Aren't naps for three year olds?

.......

....

```

Chapter Names with Timestamp (Beginning and End of chapter):

```

print([(x.data['subheading'],x.timestamp,x.timestamp_end) for x in Output.transcription.proofread.segments])

```

```

[('Canva', datetime.timedelta(microseconds=470000), datetime.timedelta(seconds=200, microseconds=210000)),

('The Economist', datetime.timedelta(seconds=200, microseconds=450000), datetime.timedelta(seconds=244, microseconds=10000)),

('The Alertness Edge', datetime.timedelta(seconds=244, microseconds=310000), datetime.timedelta(seconds=517, microseconds=10000)),

('Professional Athletes and Sleep as a Strength', datetime.timedelta(seconds=517, microseconds=130000), None),

('The Importance of Nighttime Work', datetime.timedelta(seconds=789, microseconds=890000), datetime.timedelta(seconds=1009, microseconds=550000)),

('What is the Graveyard Shift?', datetime.timedelta(seconds=1009, microseconds=910000), datetime.timedelta(seconds=1048, microseconds=670000)),

("What's the impact on shift scheduling?", datetime.timedelta(seconds=1049, microseconds=90000), datetime.timedelta(seconds=1300, microseconds=430000)),

("What's the result?", datetime.timedelta(seconds=1300, microseconds=670000), datetime.timedelta(seconds=1425, microseconds=290000)),

("What's the experiment?", datetime.timedelta(seconds=1425, microseconds=710000), datetime.timedelta(seconds=1547, microseconds=930000)),

('How the experiment was conducted', datetime.timedelta(seconds=1548, microseconds=350000), datetime.timedelta(seconds=1803, microseconds=890000)),

('What did you find?', datetime.timedelta(seconds=1804, microseconds=250000), datetime.timedelta(seconds=2003, microseconds=810000)),

('What do you think is the impact on productivity and well-being in India?', datetime.timedelta(seconds=2004, microseconds=230000), datetime.timedelta(seconds=2270, microseconds=630000)),

("What's next?", datetime.timedelta(seconds=2273, microseconds=150000), datetime.timedelta(seconds=2392, microseconds=610000))]

```

Creating a summary for each chapter:

```

chapter_names = [x.data['subheading'] for x in Output.transcription.proofread.segments]

Chapter_text = [x.output_spans[0].text for x in Output.transcription.proofread.segments]

chapter_summary = []

for text in Chapter_text:

    summary = oneai.Pipeline(steps =[oneai.skills.Summarize()]).run(text)

    chapter_summary.append(summary.summary.text)

print([x + " : " +  y for x,y in zip(chapter_names,chapter_summary)])

```

```

['Canva : Canva is a visual communications platform that gives workplaces of all sizes a unique, flexible and collaborative way to make work feel like play. Canva lets you access your slides on any browser or mobile device design, edit or present on the go.',

'The Economist : The Economist Heather Schofield: If we can help improve their sleep, it will very much improve their ability to work longer and work harder and work better. She says there are some departments that have implemented nothing policies.',

'The Alertness Edge : Reaganomics radio is a podcast that explores the hidden side of everything. Stephen dunner says sleep deprivation costs the U.S. more than $400 billion a year in economic losses.',

'Professional Athletes and Sleep as a Strength : The military launched a new field manual highlighting sleep as an important factor in maintaining operational readiness. The Olympic skier, Mikaela Schwerin, the Olympic swimmer, Ryan locate, the NBA legend, LeBron James are all famously fond of napping.',

"The Importance of Nighttime Work : An estimated one in 5 workers regularly works at night on what's sometimes called the Graveyard Shift. The World Health Organization is called night shift work, a probable carcinogen.",

'What is the Graveyard Shift? : Less experienced police officers typically tend to be on the Graveyard Shift. There is a small cohort of officers who like it who opt for the Graveyard shift.',

'What\'s the impact on shift scheduling? : There is a push now to actually factor in chronotype when shift scheduling and assigning shifts. "Fatigue management training isn\'t necessarily just beneficial for the health and wellness of officers," says Greg Slinky.',

"What's the result? : Amazon advertising is a business to business service selling global advertising space. Free economics radio is sponsored by ring has security products for every corner of your home, and you can see it all in one simple app right from your phone.",

"What's the experiment? : Half of the people in the study will be given things to help them sleep better at night. They chose Chennai in part because it has so many factors that conspire against good. The researchers outfitted everyone with a wearable sleep monitor.",

"How the experiment was conducted : The percentage of time in bed that's actually spent sleeping is called sleep efficiency. The average sleep efficiency in high income countries is 85 to 95%. The only sleep 5.5 hours, which tells you there's a sleep efficiency of about 70%.",

'What did you find? : Most experts predicted that work hours would go up and the idea was, maybe a less tired, you could spend a little bit more time at the office and so on. But instead, we find none of that.',

'What do you think is the impact on productivity and well-being in India? : Sleep quality in a cool, quiet environment is likely higher than the quality of sleep, author says. Some employers might be able to do it, but the typical low income worker in Chennai is not able to, in fact, do that.', "What's next? : Reaganomics radio is produced by stitched and run bud radio. You can get the entire archive on any podcast app if you'd like a transcript or show notes, go to Reaganomics dot com."]

```

Conclusion

Taking the measure to utilize OneAI’s Language Studio to add transcripts to your podcasts is an incredible way to boost your SEO and accessibility. After taking this step, the next step is simply to split it by Chapters, making it easier to retrieve information from certain points in the podcasts. 

OneAI offers the opportunity to utilize our software and code for your ease, making the production of your hard work all the more easy.

Explore our Language Studio's other skills and join our community to share your thoughts.

‍