Google I/O 2025 kicks off on Tuesday, and we expect the keynote to focus entirely on AI announcements. Last week, Google unveiled the top Android 16 upgrades, including the new design, better security, and Gemini replacing Google Assistant on every smart device.
Google I/O might as well be called Gemini I/O at this point, though. AI has been its biggest priority since ChatGPT arrived, and Google has been making significant progress in the field over the past few years. The newest Gemini models have been a success, with Google continuously tweaking their performance and adding new features.
For example, Google has just updated Gemini 2.5 Pro to improve its coding abilities. It also brought image generation and editing support to Gemini users. Google also included Audio Overviews in the Gemini experience, a previously NotebookLM-only feature that lets you turn research reports into interactive podcasts.
I said more than once that I’d love ChatGPT to get a similar feature to make consuming the detailed Deep Research reports from OpenAI’s models easier. I don’t know whether we’ll see such an upgrade from OpenAI anytime soon, but Gemini might get something even better than that, and the feature leaked ahead of a potential I/O 2025 reveal.
I’ve often used the ChatGPT Deep Research feature since it was released a few months ago. But I’d love an easy way to listen to the reports while I run, especially those I don’t have to read right away. That’s what Audio Overviews offer Gemini users.
I’ve also often wondered when I’ll be able to tell ChatGPT to make me videos that explain certain concepts. I ask the AI to give me graphical representations for certain ideas or make graphs for others. But that’s about it. What if the AI could create video tutorials/summaries on the spot?
It turns out that Google might be working on that sort of functionality, which might be available to Gemini users in the future as a Video Overviews feature.
Nothing has been confirmed, and Google’s I/O event kicks off this Tuesday, but TestingCatalog has found evidence that such an all-in-one product might be coming to Gemini soon.
The blog looked at Illuminate, a Google AI experiment that lets you turn content into AI-generated audio discussions. That sounds like NotebookLM technology at play, but it’s a different product. TestingCatalog says a new Illuminate version has been rolled out, packing features that are hidden from most testers.
The blog says that Illuminate might let users create Audio Overviews for all sorts of content, not just research papers. Classic books like Alice in Wonderland and The Great Gatsby might also be supported.
The new UI features experimental controls, such as an Edit button, caption toggles, and a cover image generation tool.
However, the most interesting new Illuminate feature is a section called Sparks that’s in Early Preview with the following description:
Imagine any question could be instantly transformed into a short video, 100% AI-generated.
Now, this is a reason to get excited about a Video Overview feature coming to Gemini in the near future. TestingCatalog shared several samples on X that reveal how the feature would work. You’d give the AI a question, and Illuminate will create a video complete with audio commentary that addresses the prompt.
These clips are between one and three minutes long, so they could hardly cover a large ChatGPT Deep Research report. However, they’d be perfect to explain all sorts of concepts with the help of visual cues. It would make understanding certain topics even easier, while others might be more entertaining to certain audiences.
It’s unclear what AI models the experimental project uses. The blog speculates that, due to the high quality of the videos it can put out, it might connect a video generation tool like Veo 3 to a multimodal next-gen version of Gemini, a Gmini Ultra tool that Google could always unveil at I/O 2025.
Also, the videos keep the podcast-like experience in place, with AI-generated hosts discussing the topic. Therefore, the Sparks feature could always become some sort of Video Overview feature across Gemini-powered products.
As exciting as all of that might be, we should all remember that AI tech isn’t cheap, especially the kind that lets you generate videos from scratch. The longer and the higher quality the clip is, the more expensive it might be. I wouldn’t expect Video Overviews to be as widely available as Audio Overviews anytime soon. Also, you might have to pay for it via a Gemini subscription.
But the ability to tell AI to illustrate a concept, a piece of research, or a book in video form is definitely exciting. Such tech should get better and more widely available in the future, and I can’t wait for OpenAI to bring a similar tool to ChatGPT.
While we wait for Google’s big AI announcements at I/O 2025, you can check out more Sparks samples at this link.
