AI and Podcasting: The beginning of a beautiful friendship?
by Tina Nole, Larj Media Founder and Chief Creative Officer
I’ve been going deep on AI over the past six months. Things have been changing so fast, and this is the first moment I’ve had to stop, reflect, and share. Like most everyone else, I’m overwhelmed! But I’m also inspired, optimistic and, sure, slightly freaked out.
Image courtesy of DALL-E 2 AI
For the podcast industry, AI holds a lot of promise in pre-production, postproduction and for creating podcast promotional content. I’m turning over rocks non-stop! So, this is the first of a regular post on all-things-AI in the podcast industry.
Here’s my current thought process about how we start to integrate AI into our workflow, what the present limitations are and what, I hope, the future will hold:
Podcast Pre-Production
Tools: ChatGPT
An efficient way to summarize the initial premise for a podcast or individual episodes, ChatGPT could be a useful tool to help producers discover narrative story structure.
By kickstarting the brain-storming process with feedback and suggestions, ChatGPT could help to refine and improve ideas.
ChatGPT could also be a great resource for research, however its data-sourcing only goes up to 2021 so it requires some vigilant fact-checking.
Podcast Production
Tools: Descript, Adobe Project Shasta
Editing text through platforms like Descript has been a game changer for the industry and continues to evolve. It’s already making it easier to collaborate with the whole creative team and has leveled the playing field for those who are less technically inclined.
· The upcoming changes in Descript will completely overhaul our workflow and create a more efficient way of doing some of the detailed editing work. But even now, In its current iteration, this tool can remove the “um’s” and “ahs” in one keystroke and identify and clean up pauses. The ongoing evolution of these kinds of platforms will make the initial pass through an audio edit immediate – paving the way for creative editors to layer in sound and further expand the “theater of the mind” experience for listeners.
· The ability to regenerate audio is another significant advance. By highlighting a sentence or two and having AI regenerate the audio, we now have the capability to remove background noise and/or make a statement more succinct without having to re-record or completely scrap a great piece of audio. Facing this problem used to be a trade-off between quality and cost/time – that’s no longer the case.
· Descript's Overdub lets you create a text-to-speech model of your voice or select stock voices. Instead of having a host come back into the studio to do pickups, the overdub feature allows editors to correct or add in audio with the host's voice.
· Adobe’s Project Shasta has a lot of the features Descript offers plus, remote recording, ai powered audio and more (I’ll get into that in an upcoming blog post). They also offer a feature that can turn poor sound quality into broadcast quality sound in minutes. In the near future there will be zero excuses for poor quality.
Podcast Post-Production & Discoverability
Tools: CapSho, Swell AI, Momento
The ability to create post-production assets using the AI transcript has opened up a whole new world. This ability allows us to build an entire platform around a podcast in the most efficient way possible. In minutes these tools spit out a host of promotional assets including and not limited to:
Episode Titles
Episode Descriptions
Show Notes
Facebook / Instagram Promotional, Engagement & Educational Captions
LinkedIn, posts and articles
Tweets of all flavors
TikTok posts
Emails
Blog Posts
Quotes curated from the episode
YouTube Descriptions
Transcripts
Episode summaries timestamped
Here’s a quick review of the tools that are currently on my virtual workbench (more coming on the scene everyday):
Capsho: Great interface, easy to use and doesn’t have complicated ChatGPT plug-ins (not yet anyway). Capsho simply spits out 4 options for each piece of content: 4 titles, 4 summaries, 4 social media posts etc. It's an efficient way to get started because it quickly rids you of the blank page syndrome and though you can edit the content there isn’t an option to ask the AI to regenerate on== its output. You only get to choose from the 4 options it generates. I like this because it saves me from the rabbit hole of endlessly perfecting.
Momento: Unlike Capsho Momento uses more ChatGPT functionality allowing you to be more creative as you build out show notes, social posts etc. Momento also has the ability to create short videos from audio. This makes for quick and easy audiograms using key moments from your podcast episode. One thing I love about Momento is that it’s very user-friendly and has a whole library of prompts to play with and essentially helps you learn how to use it by suggesting various prompts while also allowing you to create your own. From there, you can publish social content directly from the tool, which is a nice feature.
On the down side, Momento has a lengthy URL that makes it hard (studio.momento.fm) to find and the interface is pretty rugged, so this tool requires a more seasoned AI user to get the most out of it, though in the end it is more robust than the other tools I’m playing with.
SwellAI: Very similar to Momento but without video capability and much smoother user experience. Swell offers content templates you can seamlessly integrate into the post-production process with customizable content blocks and a ChatGPT tool.
Something I haven’t seen anywhere else is SwellAI’s feature that can be integrated into a website, allowing listeners to ask questions and engage with a podcast through AI. Swell provides an embed code so you can use this tool on the podcast’s landing page allowing for more engagement. Swell also has a feature in beta right now that generates and exports multiple blog posts from a list of keywords in your podcast and once created, Swell will write an SEO optimized blog post for each keyword. That will go a long way for discoverability.
Here’s what’s on my AI podcast wish list: An all-in-one solution that supports the entire production workflow from pre-production to promotion. Seems Descript is most primed for it at this time. Here’s what I’m dreaming of:
Pre-production assistance:
Plug raw audio and have the AI summarize the content to help producers with story arcs
Chat GPT plugin for research (that’s current to within a day!)
Audio quality processing (like Adobe Mic Check)
Text-to-audio editing: Everything Descript does now, including Overdub and what they revealed in their recent announcement.
Post-Production asset creation
Automatic publishing to all podcast platforms so essentially marries Megaphone with Descript and Memento or Swell AI etc.
All the tools I’ve mentioned come with a lot of bugs, which for now is simply part of the experience - mostly because of the rapidly changing features.
I’m excited about this budding relationship and look forward to what’s next. So much is on the horizon, in fact Capsho will celebrate its 1-year anniversary this week when they’ll release more features and Descript announced last week some incredible advancements – here's that video. More to come!
Coming up:
I’ll take a look at this new tool that appears to be a one stop shop catering to the DIY podcast creator.
Adobe’s Project Shasta
A review of Descript & Capsho’s new offerings