I tested this new AI podcast tool to see if it can beat NotebookLM – here’s how it did

The Speechify text-to-speech app enables its over 50 million users worldwide to convert any text, including documents, articles, PDFs, and images into audio, with over 200 voices to choose from. Now, the company is delving into a new type of audio: AI-generated podcasts. 

Also: I finally gave NotebookLM my full attention – and it really is a total game changer

Starting today, Speechify users will be able to turn any content into a “lecture-style” podcast. They’ll also get access to a slew of other AI-powered features, including updates to AI Summaries and AI-generated quizzes and a new AI recap feature, and perhaps the biggest, AI podcasts. 

Keep reading to find out how the new AI podcast feature works and how it compares to viral tools such as Google’s NotebookLM Audio Overviews AI podcast feature or ElevenLabs’ AI podcast option.

Who is Speechify for? 

Speechify is geared toward people who either have a reading impediment or just prefer voice content. The new releases are positioned as study aids and back-to-school tools for students. 

How does the AI-powered podcast work?

On the Speechify platform, you have always been able to upload your own materials for it to read them out loud for you. Now, those materials can be used to create a podcast. One feature that differentiates it from competitors like NotebookLM is that on the Speechify platform, you can choose the style of “show” that you want to listen to, with options including a standard podcast, debate, late-night with notes of comedy, and lecture.

Instead of uploading your own text, you can also enter a text prompt of a topic you’re interested in, and Speechify will generate a podcast about it for you, another feature that sets it apart from NotebookLM. Speechify does not delineate what source material it uses to generate a podcast from scratch, but told me that this feature is powered by its proprietary large language model (LLM), which is comprised of open and closed source models. The company says it runs constant evaluations to make it better, but since it pulls from the web and the podcast doesn’t include citations, it’s wise to verify any statements or claims, especially if you’re conducting research for school. 

How it worked for me (and how to access) 

Using the feature was intuitive. The first step was to click on “AI Podcast.” Once you do, you are presented with the four options and a description of each. For example, the podcast mode is described as “neutral tone, two speakers,” the late-night show is described as “notes of comedy, three speakers,” the debate is described as “argumentative, three speakers,” and the lecture is described as “educational, two speakers.” 

Also: This new AI podcast generator offers 32 languages and dozens of voices – for free

Once you click on your mode of choice, you can either click on the “describe with text” or “select file” option to create the podcast. For the sake of this experiment, I uploaded a PDF of a ZDNET article I wrote. After you upload a file, you can try the traditional text-to-audio feature, in which a voice reads the content to you word for word (ElevenLabs offers this feature as well). This is pretty neat in itself; I had Gwyneth Paltrow read my story to me. 

I chose the AI podcast option in the podcast mode, which took one minute and 25 seconds to create. In the world of near-instantaneous AI, I was surprised that it took a bit more than a minute, but all things considered, a minute is still pretty quick –NotebookLM’s process takes a few minutes. 

The resulting podcast was great. It was about five minutes (half as short as what NotebookLM currently offers), accurate to what I wrote, and sounded like a human conversation. A unique feature I found helpful was a transcript of the conversation that you can easily follow along with as the words are highlighted as they are read, like when singing karaoke. Other helpful features let you control the speed and easily revisit the original text, which is displayed on a banner at the bottom of the app page.  

Also: Why AI chatbots make bad teachers – and how teachers can exploit that weakness

I then tried it again with the late-night show and debate modes, as the nuances involved in carrying those out sounded the most intriguing to me. The late-night show was surprisingly better than I anticipated, as I was nervous the three-person chat would be overwhelming. Instead, there was decent banter and hand-off between the three protagonists that made it much more enjoyable to listen to than the standard. The humor added vivid analogies and metaphors that worked really well to breathe live into the content without changing its meaning — likely especially helpful for denser topics. 

The debate mode gave me one speaker’s intonation that I could not bear to listen to because of the chipper, high intonations that were unnatural to what was being said — but this isn’t a deal breaker to me since you have the option to pick which mode you would prefer, and someone may be a fan of the overly cheery. 

Speechify Premium users can try the AI Podcast feature on iOS today; web and Android are still on the way, though Speechify did not specify a release date. 

Other features 

Speechify has already been using AI to fuel other features on the platform, including generating summaries and quizzes, before moving into podcasts. Those features also got an upgrade that allows users to create summaries and quizzes from specific chapters within their reading rather than the entire text. 

Also: Claude Sonnet’s memory gets a big boost with 1M tokens of context

AI Recap is a new AI feature that allows users to click “recap” to get a summary of what they just read, similar to the refresher you see before watching the next episode of a TV show. 

Speechify’s Chrome Extension got some updates, including the addition of AI Summaries. This lets users summarize any text directly in their browser’s side panel. Users can also hone in on part of the text by using the “ask AI” feature. Speechify shares its Chrome Extension was also upgraded to be faster, loading almost instantly without slowing down browsing. 

Some non-AI additions include the ability to take notes, highlight text, bookmark readings, search in documents, share files, and upgrade voices that sound more conversational. Speechify is also enabling a filtering experience that makes it easier for users to find just the right voice for them. 

Also: Don’t fall for AI-powered disinformation attacks online – here’s how to stay sharp

Lastly, in the back-to-school spirit, Speechify also added over 70,000 free books to Speechify, including classic literature, OpenStax’s catalog of free textbooks, and 130,000 textbooks from other publishers such as Taylor & Francis and O’Reilly that users can access directly in the platform.