One of the strengths of podcasting is the close connection that listeners can feel to the host whose voice is in their head. But now, tools like Jellypod and NotebookLM are taking the human podcaster out of the equation and creating podcasts with AI.
Google’s NotebookLM has been a major talking point lately, mostly due to its Audio Overviews. This feature allows users to turn any content into dynamic AI-hosted podcasts. But while NotebookLM dazzles with its remarkably realistic voices, it leaves room for improvement in customization. Enter Jellypod—an AI podcast creator that might just be the solution for podcasters who want to publish real episodes. Jellypod is currently in limited release – new users can join the waitlist.
NotebookLM: A Groundbreaking Start
Google’s NotebookLM deserves the attention it has received. With two lifelike AI hosts, the platform generates conversations about user-provided content that sound amazingly human. However, the current version of NotebookLM has its limitations.
For one, users are stuck with the same pair of hosts—no solo options, no changes in age, gender, accent or language. While you can direct the hosts to focus on a topic or audience, customization is minimal. Want to predefine episode length or tweak the content it creates? You’re out of luck. These constraints make NotebookLM more of a demonstration platform than a production tool, at least for the moment.
How Jellypod Raises the Bar
Jellypod steps in where NotebookLM falls short. Designed as a comprehensive podcast creation tool, it offers podcasters significantly more flexibility:
Jellypod’s Customizable Hosts
Jellypod allows creators to set the number of hosts, which allows both solo podcasts and multi-host formats. Its library of professional voices spans accents, ages, and styles, from a “deep, theatrical middle-aged American male voice” to a “young Australian female voice.” You can even clone your own voice or craft a unique host persona by defining traits and characteristics.
Content Control in Jellypod
Jellypod lets users add up to three URLs or input text directly, and creates an outline that breaks down content into “chapters” with key points. (This is far fewer than NotebookLM’s 50 sources, but should be sufficient for most situations.) In contrast to NotebookLM, creators can edit these outlines extensively—adding, removing, or tweaking chapters and bullet points. While the final transcript can’t be directly edited, the platform lets you iterate the outline editing.
Jellypod’s Multilingual Capabilities
In today’s globalized world, the ability to create content in multiple languages is no longer a luxury. Jellypod delivers on this front, making it an appealing option for creators with diverse audiences.
Podcast Publishing in Jellypod
For users who intend to publish their audio content in podcast format, Jellypod has a feature that lets one append a custom intro to each episode. They also set up your podcast feed so that you can distribute content easily to Spotify and other podcast platforms.
Creating with Jellypod: A Hands-On Experience
Using Jellypod is straightforward: input content, refine the outline, and generate the audio. For example, a recent test produced an episode outline with neatly categorized chapters, each running two to four minutes. Once the transcript was finalized, the audio was ready in minutes.
Though Jellypod’s AI-generated voices are competent, they lack some of the natural nuances of NotebookLM’s hosts. Despite the plethora of voice customization options, Jellypod’s hosts can still sound a bit monotone or mechanical. Yet, the trade-off is clear: Jellypod empowers creators with flexibility and production-ready tools.
Here’s a sample clip from a five-minute podcast created using Jellypod. It features two hosts—a voice clone of me, Roger Dooley, and “Nigel Watson,” a host I created using Jellypod’s selection of professional voices.
What do you think? My voice clone is recognizably me, albeit imperfect. At times, the voices seem quite human. At other times, they sound robotic or just strange. The places where one host interrupts or talks over the other seem a bit random. Jellypod hasn’t quite crossed the uncanny valley.
Which is Better: NotebookLM or Jellypod?
In its current form, Google’s NotebookLM is an impressive showcase of AI’s potential. Its hosts sound human, with lifelike intonation. They use occasional “ums,” “likes,” hesitations, and other disfluencies that mimic real speech. They even ad lib comments from their own “experience.” One use case where NotebookLM excels is taking a batch of content – articles, books, research papers – and creating easily digested audio summaries.
In addition, NotebookLM is a far more versatile AI model. Its impressive Audio Overview feature is just one of the many ways users can interact with their content.
But for creators who need a practical AI podcast production tool, Jellypod’s array of customization options makes it the more compelling choice. The ability to use just one host is huge for creators who want to turn their audio into video with human avatars using a tool like HeyGen. Audio with two hosts requires tedious editing and yields imperfect results when the hosts talk at the same time. (Yes, the NotebookLM hosts interrupt each other, insert the occasional “uh-huh,” etc. – realistic, but editing gets trickier.)
With Jellypod, podcasters can tailor their episodes to specific audiences, formats, and languages. By editing the “chapters” they can exert considerable control over the generated content. While it’s not perfect—voice realism could improve, and transcript editing before generating audio would be nice—Jellypod is a solid option for creators who want to publish their AI-generated audio in podcast format.
NotebookLM, Jellypod and the Future of AI Podcasting
Both tools are evolving rapidly. Google is likely to expand NotebookLM’s capabilities, and Jellypod will surely continue to refine its voice technology. For podcasters and other creators, these advances will offer exciting possibilities, from hyper-personalized content to seamless production workflows.
As a podcaster, I find these developments to be both scary and exciting. Authors are already releasing audiobooks with their cloned voice, and listeners apparently think they sound just fine. How long will it be before podcasters discard their studio gear and just let AI write the script and record the audio?
Interview-style podcasts seem safe for the moment, but perhaps the day of, “My AI will talk to your AI,” isn’t that far off. Could Jellypod’s creations like my Nigel Watson eventually become more engaging than many humans? Scary and exciting, indeed.