The latest advancement in artificial intelligence has made producing podcasts easier than ever, while blurring the line between human and AI voices. Google’s NotebookLM, a tool originally designed for research and summarisation, now features a text-to-voice function that transforms written text into a lifelike podcast conversation between two AI-generated hosts.
TechRadar senior editor Graham Barlow tested the tool by creating an eight-minute podcast from a blog post, describing the outcome as “so natural and realistic” that it felt indistinguishable from human conversation. “The world simply wasn’t the same anymore,” he said, as he questioned his ability to discern real voices from synthetic ones.
A Game Changer with Controversy
Originally launched as a research tool to summarise notes, NotebookLM is now seen as a revolutionary AI-driven audio companion. Using Google’s Gemini 1.5 model, the new feature brings AI hosts to life by having them summarise material, make connections between topics, and engage in seemingly authentic back-and-forth banter.
But not everyone is embracing this technology. ZDNET’s David Gewirtz remarked that the tool’s ability to mimic human voices so convincingly was unsettling, calling it “the devil’s work.” As a long-time content creator, Gewirtz admitted feeling the pressure from AI-driven content’s rise, as the quality of NotebookLM’s “voice fidelity” and the hosts’ organic “banter” brought a chilling realism to AI-generated audio.
Democratising Audio Content
Despite the concerns, some see this development as a game-changer. Gewirtz pointed out that while it may cost Google billions to develop the technology, creating an AI podcast takes only moments and is virtually free for users. NotebookLM’s feature dramatically lowers the barriers to entry for podcasting, potentially allowing anyone to generate high-quality audio content without traditional resources.
Currently, the AI-generated voices are limited to a standard male and female voice with American accents, but future versions may offer options for customising speakers’ accents, styles, and even the tone of voice. This opens up possibilities for more realistic, personalised experiences that could further transform the way audiences engage with digital audio.
AI-Created Podcasts: The Case of ‘Pager Protocol’
NotebookLM isn’t alone in the human-free podcasting space. Recently, just hours after explosions hit Lebanon, a new AI-generated podcast, Pager Protocol, appeared on streaming platforms. This fictional podcast series by Caloroga Shark Media used generative AI tool Claude to develop a storyline based on the real-world events, which was then refined and voiced by AI narrators.
While this rapid-response content production illustrates AI’s potential to produce audio at breakneck speed, some experts worry it could undermine the core appeal of podcasts. Jason Saldanha, COO of PRX, a nonprofit digital radio company, argues that podcasts thrive on a unique, one-to-one connection between hosts and listeners. “Flooding the market with content to get the lowest level of engagement is not a long-term strategy,” he cautioned. “It’s short-sighted and ultimately harmful.”
What Lies Ahead for AI in Podcasting?
As AI advances in creating increasingly realistic voices and natural conversation flows, questions around authenticity and audience trust are likely to grow. AI-generated hosts may be convenient, but in the long run, the unique bond between human hosts and listeners could be irreplaceable.
While AI brings powerful tools to the industry, it’s clear that ethical considerations and long-term impacts on listener engagement will be crucial in shaping the future of podcasting. The challenge will be balancing the ease of AI with the authenticity that listeners crave, as the industry navigates this new frontier.
