Rounding up Google I/O 2025: 5 significant announcements

AI unsurprisingly dominated the latest round of Google's annual I/O developer conference. Here are some of the highlights, including an unexpected new AI filmmaking app.

Google doubled down on all things AI at its Google I/O event, with more 'traditional' technologies such as Android taking a bit of a back seat. Obviously it smells weakness in its great rival in the mobile operating space Apple, whose AI strategy has been patchy to put it politely. But there's also a huge arms race going on. It seems daft to talk about first mover advantage given that ChatGPT is now two and a half years old, but the feeling is that no one has yet made a compelling consumer case for it. Someone needs to soon though, as a lot of money is being spent here...

So yeah, AI is in everything. With that in mind, here's some of the most interesting stuff that the company announced at the event. For a full rundown, head to https://io.google/2025.

Flow AI filmmaking app

Flow is a nested part of an overall overhaul of the Google AI imaging pipeline, namely a new Veo 3 video generation model and a new image generation model, Imagen 4. It does all the usual things you would expect from an AI of this ilk in this day and age; using prompts including text-to-image to assemble eight-second clips. 

Veo 3 & Veo 2 upgrades

We're into the third generation of video generators now, and the sophistication is ramping up. As with Sora and other rivals, Veo 3 can create longer clips, but where Veo differs for the moment is that it also adds synced audio. And that includes dialogue as the above clip shows. Veo 2 meanwhile has been graced with new tools including camera controls and object removal.

Imagen 4

Gemini looks to have pulled ahead of its rivals with the latest Imagen 4 model. For a start, it's boosted revolutions to 2K, while Google also says it's boosted its photorealistic properties  which now include "remarkable clarity in fine details like intricate fabrics, water droplets and animal fur, and excels in both photorealistic and abstract styles." Spelling and typography have also been ramped up, plugging one of the significant holes in AI image generators' current arsenals.

Project Aura

Google Glass was a now canned project that it's reasonable to say was ahead of its time. But with all things XR suddenly hot again now that the technology has evolved to meet expectations, the company is trying again. Google and Xreal have unveiled a strategic partnership for a new Android XR device called Project Aura which will effectively embed two mini screens in a chunky old pair of sunglasses. Details are few and far between, but with Apple apparently working on similar tech, expect progress and announcements to accelerate soon.

Google Starline Beam

The newly remonickered Google Beam is designed to make video calls more realistic by using six cameras and a light field screen to generate a 3D image. No VR goggles are required on the one hand, but all parties on the conversation do need to be sitting in a special booth on the other. Deloitte, Duolingo, and Salesforce have already said that they will add HP-made Google Beam systems to their offices this year. Feels like a cul-de-sac at the moment, so will be interesting to gauge real world impact.