You can’t avoid the hype around Generative AI right now. Artificial Intelligence has been used by creatives for years to enhance their work, but it’s made new breakthroughs in the last 12 months. 2022 saw the mass adoption of generative AI image services like Dall⦁E 2 and Midjourney, and the introduction of the revolutionary text-generative AI bot ChatGPT. The goalposts have been moved.
From the mingling of awe, excitement, fear and confusion that’s followed these developments, it’s clear we’re at a major turning point in the use of AI technology, on the scale of the desktop computing and internet revolutions before it.
Before we get into using AI as a tool for podcasting, however, let’s address those fears. All disruptive technologies bring unprecedented issues, both ethically and legally; just think of publishers’ concerns about the impact of Google on their business models, or the bombshell of Napster’s file sharing service in the music industry.
Likewise, it’s fair to say that artists and creators have very real concerns about the use of AI tools discussed here. At the time of writing, there are class action lawsuits in motion, claiming copyright violations that impact software creators, artists, writers and musicians.
On the other hand, these tools can be used ethically, in ways that don’t contravene copyright laws or compromise fair use. When used appropriately, AI tools can be incredibly powerful assets that help podcasters create better content more efficiently - as long as they’re approached with caution and consideration.
In this article we’ll look at how you can use AI to help your podcasting. The tools suggested are the most relevant at the time of writing, but we encourage you to do your own research as new AI platforms are emerging all the time.
Research, planning and podcast creation
Generative AI algorithms create new content, based on large language and image models. Podcasters need to make countless creative decisions and tools like ChatGPT can help you come up with ideas at speed. For example, you can use generative AI to suggest interview questions for guests, or generate ideas for segments, recurring features and episode titles, as well as quickly write first drafts of scripted segments, podcast show notes, social media posts and any other copy you need.
The likes of ChatGPT and Jasper AI can be used to plan your podcast episodes too. Because most generative AI tools crawl the web for information to build their models, they’ll be able to call on specific information about not just your podcast, but your guests as well, which can allow them to build useful things like outlines for episodes. For example, you can ask it to write an outline for an episode of your podcast based on an interview with a specific person around a specific topic. You can also generate ideas for formats, such as solo episodes and interviews, and even get ideas for potential guests based on parameters such as location and specialism.
If you’re covering a new topic that you don’t know much about, AI tools can also give you useful background information. For example, you could ask ChatGPT for a brief summary of a sporting personality’s career highlights, saving you hours of trawling through Google search results and other online sources.
Beyond creating content from scratch, AI is useful for transmuting existing assets, and AI-powered platforms like Descript can save you a lot of time and effort in the production process. The software transcribes your recording into text, and you can then manipulate the text to edit your audio, for example by cutting unnecessary segments and chopping out filler words.
Podcast promotion and marketing
Posting a written transcript of your podcast on your website is good for accessibility, allowing those with hearing impairments to access your content more easily, but it can also make your podcast more discoverable by Google and other search engines. Manually typing it up is a tedious chore, however, so transcription tools like Descript, Otter.ai or Speechtext.ai can be used to generate text from your podcast audio. This is useful for grabbing quotes for promotional content too - just be sure to double-check it’s been transcribed properly with no errors.
Tools like Descript and ElevenLabs also enable you to generate speech from text you type, to drop into your audio production. They use realistic sounding presets, and can even clone your own voice - although both platforms insist you have full rights for any voice cloning you generate, for obvious reasons. AI-powered post-production tools like Auphonic, meanwhile, analyse your audio and make recommendations for improvements, such as removing background noise and adjusting volumes. It can suggest sound effects and music to enhance your recordings too.
Many podcast management platforms also use AI to enhance podcast distribution. For example, AI-powered tools like Podcast.co and Audioburst analyse audio content and categorise it into searchable short-form audio and video clips, which you can use to promote your podcast.
Also useful for podcast promotion are image generation tools like Dall⦁E 2 and Midjourney generate extraordinarily high quality images in a range of styles that can be used for supporting content, such as social media and blog posts. Prompts should be based on descriptions of the image’s contents, and any desired stylistic requirements, such as “a charcoal sketch of an environmental activist cuddling a baby polar bear”.
Using AI safely
As illustrated above, generative AI has provided some incredibly useful tools - but there are some important caveats and conditions to bear in mind for podcasters who want to incorporate them into their workflows. Because they’re often trained on models that use data scraped from a huge amount of publicly-accessible sources on the internet, you should always approach their output with a degree of skepticism. While they can be a great way to spark ideas and start conversations, the often-patchy factual accuracy of generative AI models means the results shouldn’t be taken at face value or used directly in any public-facing forum.
Some inaccuracies are inevitable, so always double check the factual details of any results before you build them into your script. It’s absolutely imperative that you take time to review, fact check and edit any content produced by AI tools, ensuring generated content is accurate, original and written in your tone of voice.
Image generation tools like Midjourney come with even greater caveats, as many image generation models are trained by automatically scraping and analysing the work of artists without their knowledge or consent. You should never use AI to recreate artwork in the style of an existing creator; we also recommend running a Google reverse image search on any images you create with AI, to make sure you haven’t inadvertently plagiarised another artist.
In addition, AI still can’t match the talent and creativity of a human artist, so it’s always preferable to work with a real designer or illustrator for your projects wherever possible. As a general rule, stick to using AI for non-critical art assets such as images accompanying social posts, rather than for headline images, episode artwork or anything with a long shelf-life. Finally, if AI has been used to help create something public-facing, you should include a disclosure indicating how it has been used - such as the one at the bottom of this article.
AI is here to stay
With all that being said, AI tools can be a fantastic timesaver, allowing podcasters to extend and speed up their capabilities. There are already many professionals using AI to improve their processes and create higher quality podcasts. For example, Joanna Penn, host of popular podcast the Creative Penn, uses AI at all stages of her podcast production process.
“I edit my podcast interviews in Descript,” says Joanna, “using the text function to highlight words to remove and delete as appropriate. I use Otter.ai for transcription as I find my British female accent alongside an American accent, or another nationality, doesn't quite cut it with Descript's transcription. And I use Auphonic.com for leveling and post-production, which uses AI in its audio algorithms.”
It’s still early days for AI, so keep an eye on developments, as this technology is sure to develop in even more unpredictable ways. As Andrew Ng, Chief Scientist at Baidu Research and Stanford professor, says: “‘Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don’t think AI will transform in the next several years.”
The main image for this article was based on an image produced by Dall⦁E 2.
