AI creates new narrative podcast in just 24 hours

Production agency used machine learning tools in writing, voice acting, sound design, artwork, and more

Audio content agency Distorted has released a new scripted narrative series, generated entirely by artificial intelligence in just 24 hours.

The twice weekly series, titled Synthetic Stories, will cover a range of genres and was created by the Distorted team as an experiment to test if AI has the capability to create an entire podcast from beginning to end using the current technology available online. The idea was conceptualised on a Wednesday and by Thursday evening, the team had already completed the project, with the first episode launching on 15 March across all platforms.

“We started to wonder how far we could take this, how good is AI right now? Could we make a podcast series entirely created by AI using only the tools and the websites that anyone could sign up for?” said Distorted director Andi Durrant. “To be totally honest, what happened over the next 24 hours totally blew our minds.”

According to the Distorted team, ChatGPT - an OpenAI chatbot which answers prompts with detailed responses - was used to create the written aspects of the series. The team first started with creating the script, which they based on themes of AI and podcasting, using keywords and phrases like “dark twists” and “end on a cliffhanger”. ChatGPT was also used to come up with the name of the series, and a press release that was sent out to media. 

Midjourney, a website that created AI-generated images using prompts, was used to create examples of podcast artwork using similar keywords to the script, which the team then expanded on. The music was also then generated on Sounddraw, an AI music generator, also using prompts such as “mysterious” and “suspenseful”. 

Lastly, the voice in the podcast was generated via ElevenLabs, a speech synthesis and text-to-speech software, and used Distorted marketing manager Becky Lamb-Pritchard’s voice as well as links from Durrant’s old radio show as templates to create two entirely new voices. 

AI first started to make an impact in the podcasting world in October 2022, when a podcast episode featuring a deep-fake interview between Joe Rogan and the late Steve Jobs - created as a showcast by AI text-to-speech company - went viral. Since then, AI tools have quickly continued to develop and become more accessible to users.

At a panel discussion during the Campaign and PodPod Radio and Audio Summit in January this year, the speakers - Global DAX commercial director Ryan Rummery, Mars UK media and content director Richard Mashiter, and Loud Parade co-founder Rahmon Agbaje - were asked about how valuable human voices in podcasting will continue to be against the future of AI. 

Both Mashiter and Rummery agreed that having a face behind the voice that can build a human connection with an audience outside the podcast is key, and will always be vital for brands and listeners despite developments in AI. While Rummery agreed with that, he also added that there is a way for podcasters to still benefit from AI.

“What you can do with AI is dialect, tones, but the heart of it is the talent and voice - not the computer,” said Rummery. “There is a way for them to co-exist going into the future.”