Joe Rogan meets Steve Jobs in an AI-generated podcast.
What’s new: For the debut episode of a new podcast series, Play.ht synthesized a 19-minute interview between the rock-star podcaster and late Apple CEO. You can hear it here and propose computer-generated participants in future episodes here.
How it works: The Dubai-based startup created the episode using text generation and voice cloning.
- Play.ht generated the script using an unnamed natural language model that it fine-tuned on Jobs’ biography, interviews, and other sources.
- It rendered the transcript into audio using proprietary synthetic voices trained on audio recordings of each speaker. Play.ht’s voice editor synthesizes voices in over 120 languages with phonetic control over pronunciation.
- The production is the first in a series called Podcast.ai. The public can propose meetings of the virtual minds for future episodes.
Behind the news: Rogan was also the subject of an early experiment in voice cloning. In 2019, Toronto-based Dessa released ersatz Rogan audio clips — the first of a parade of fake celebrity voices.
- Earlier this year, James Earl Jones, the voice of Darth Vader, signed a deal that permits Disney to recreate the Star Wars villain’s speech using technology from Ukrainian startup ReSpeecher.
- Two documentary filmmakers separately generated vocal facsimiles of deceased celebrity chef Anthony Bourdain and iconic artist Andy Warhol. The Bourdain imitation sparked controversy when his widow revealed that she had not given the filmmaker permission to recreate her husband’s voice.
Why it matters: The declamation is occasionally stilted and the script meandering (with occasional lapses into incoherence), but the rapid progress of generative audio combined with the entertainment world’s appetite for novelty suggests that satisfying synthetic productions may not be far off.
We’re thinking: How long before we can produce Heroes of Deep Learning without actually talking with any of the heroes of deep learning?