Spotify is looking to leverage their AI technology to take podcast ads to the next level. Reports from The Ringer founder Bill Simmons indicate that the streaming service is working on a host-read ad system powered by AI that would sound like the real host without them actually having to record and read the ad copy. This would enable podcasters to geo-target ads, create ads in different languages, and provide many other opportunities.
The technology in question is learned from Spotify’s head of Cultural Partnerships and podcast host Xavier “X” Jernigan’s voice recordings, including different cadences and emotion. It then feeds this information into an AI model to create and build the AI voice, however Spotify has not shared details on the timelines or process. The current AI DJ feature is powered by IP from the company’s 2022 acquisition of Sonict combined with OpenAI technology.
The company has a few hundred people working on aspects of personalization and machine learning, where they are studying larger language models, generative voice, and more. Details regarding other potential AI voice models in the future was teased by Jernigan himself.
Overall, while it is unclear whether or not this AI-based host-read ad system will make it to the platform, Spotify has made it clear that they are dedicating time to researching and developing new applications for AI in their system, potentially opening up a variety of new possibilities in the near future.
