How podcast networks are testing AI tools for faster translation, ad sales

For translation

iHeartMedia is testing a range of generative AI tools to translate some of its podcast show archive into different languages to reach listeners in international markets, said Conal Byrne, CEO of the iHeartMedia Digital Audio Group. iHeartMedia declined to share which AI companies they were using to test the translations. The technology will be used to transcribe, translate and then voice the translations. An iHeartMedia spokesperson said the company is working with podcast producers to test the quality of the translations in both AI-generated voices and the original hosts’ voices.

The translated shows will give podcast networks “a shot to globalize a medium that has so far largely been a U.S.-centric medium,” Byrne said. “I think this may be our best shot in a while for the podcast industry to globalize — to actually serve up content in-country, in-language around the world.”

By the end of this quarter, iHeartMedia plans to debut five to 10 shows translated into languages like Spanish, French, German and Italian, Byrne said. Shows will likely be evergreen, general interest shows and limited series shows, he added, but he declined to share the names of the shows. Byrne also declined to comment on whether these shows would contain disclaimers that they were generated in part by artificial intelligence technology.

Spotify is also testing generative AI to translate podcast audio. In September, the company announced it was launching a pilot program with a few podcasters to test AI-generated voice translations into other languages.

“While this type of technology definitely brings challenges, our focus is on helping creators connect with audiences, monetize their art and build careers,” Spotify’s Brian Berner, global head of advertising sales and partnerships, said in an email.

For ad sales

Since adding AI tools to Acast’s self-serve advertising platform last summer, nearly 40% of the podcasters who were getting requests from advertisers had previously never worked directly with advertisers before, said Acast’s chief product officer Matt MacDonald.

Called Collections+, the AI-powered tool pulls podcast and listener data from different sources (including from podcast database Podchaser, which Acast acquired in 2022) to automatically group podcasts into contextual categories. This allows Acast to package up smaller shows from Acast’s catalog of podcasts to meet advertisers’ briefs, MacDonald explained.

Acast chief business officer Greg Glenday said the company is working with one national advertiser — whom he declined to name — on a multi-million-dollar campaign. The advertiser named 11 shows it wanted to buy in order to reach a particular audience. Glenday said his team used the AI tech to show the advertiser what it would look like to buy 52 shows that reached the same target demographic, but reach more scale and at lower CPMs, and is providing third-party measurement research for the campaign. The advertiser’s budget did not change as a result, he said.

“We can remove the manual labor [of reaching 52 shows to] remove all of the obstacles for that brand and that buyer. Then all of a sudden [AI] is a real media reach vehicle and not just this fun, new thing,” Glenday said. “It spreads the revenue out across more inventory and makes for a much better business.”

During iHeartMedia’s third quarter earnings call with shareholders on Nov. 9, CEO Bob Pittman said the company was providing some of its sellers with AI-enhanced tools to help them prospect and communicate with podcast clients, as well as using AI tools for dynamic ad insertion to improve messaging and voice for targeted demographics.

For production assistance

Generative AI tools can help podcasters in the production phase by assisting with tasks like researching, scripting, editing and publishing content, Byrne said. iHeartMedia is working with podcast producers to test AI tools like Jasper and Microsoft Copilot as research and writing assistants, with human oversight. While those tests are “more on the slow roll… that stuff has moved from being theoretical, nice-to-have to this quarter, it’s going to get really real,” Byrne said.

Some of the tests for Acast’s content productions are also in the early phases, MacDonald said. One use case the company is looking at is using generative AI tools to help podcast creators outline production schedules over a set period of time. Acast is also interested in using predictive AI models to evaluate trends in emerging content and podcast listener demographics to inform business decisions, such as which podcasts they add to their network, Glenday said.

But not for all things podcasting

Not all podcast networks are diving in head first. Gina Garrubbo, president and CEO of NPR subsidiary National Public Media, said AI will be a “big deal” in podcasting this year in areas like content creation, commercial messaging creation and audience targeting, but she noted that NPR is not “ready” to discuss how the technology would apply to NPR yet. Garrubbo did confirm NPR was not testing the technology yet for those use cases.

At the end of the day, the execs who are experimenting with this technology were quick to say that generative AI tools can’t replace human talent – especially hosts. 

“We’re cautious about saying that AI can do things that unique, incredible human talent can do today. But as a sort of a personal assistant if you will, to every single producer or executive producer we have – that’s the vein we’re interested in exploring,” Byrne said. “And [we’re] not overstating it – not saying that AI is going to completely invert and disrupt all of media. It’s not.”

For now, Acast is not interested in testing AI tools for ad creation or replacing a podcast host, Glenday said. “I just can’t imagine listening to a virtual voice review movies once a week,” he said.