Google’s AI podcasting tool called Audio Overview could become a surprisingly viral hit, according to an MIT Technology Review report.
The tool was launched in this year’s mid-September as a part of NotebookLM, a one year-old AI-powered research assistant, powered by the Google Gemini 1.5 AI model.
Also Read: Paytm’s CTO gets a new role called ‘AI Fellow,’ Deependra Singh Rathore replaces him
It allows people to upload links, videos, PDFs, and texts, and then ask questions regarding the content to get short summaries.
What’s more is that it can generate a podcast named Deep Dive based on the uploaded content.
This features a male and a female voice discussing about the content uploaded, which is extremely realistic, and even features human-like phrases such as “Man” and “Wow” and “Oh right” and “Hold on, let me get this right,” according to the report, which added that the hosts (The voices) even interrupt each other in a realistic fashion.
The report cites Andrej Karpathy, a member of OpenAI’s founding team as well as a former director of AI at Tesla as saying on X that Deep Dive is now his favorite podcast, with him creating a podcast series called Histories of Mysteries with topics researched using ChatGPT, Claude, and Google, and Wikipedia as source material.
Also Read: Zomato to grant nearly 12 million employee stock options worth around ₹330 crore
How to use Google’s Audio Overview to generate AI podcasts?
Its very easy to use as well. All you have to do is go to NotebookLM and create a new notebook, add a source in the form of any PDF document, MP3 audio, or even a video file. Yu can even paste a link to a website. Clicking on “Generate” in the Audio Overview feature in the top-right corner will generate it in a few minutes and afterwards, you can download it or share a link.
However, it is not entirely flawless as the report finds certain moments it hallucinated as well. Hallucinations in an AI context refers to moments when an AI spits out incorrect information.