Google’s annual I/O developer convention kicked off on Tuesday, Could 20, with a gap keynote that was principally AI all the best way.
In a protracted stream of bulletins, the search large drew the curtains again on the brand new AI merchandise it has been engaged on over the previous few months, together with an AI device for filmmaking, an asynchronous AI coding agent, an AI-first 3D video communication platform, and extra.
It launched upgraded variations of current AI fashions and instruments equivalent to Gemini 2.5 Flash and Professional, Imagen 4, Veo 3, and Lyria 2, in addition to new updates to AI Mode in Search, Deep Analysis, Canvas, Gmail, Google Meet, and so forth.
“Extra intelligence is on the market, for everybody, in every single place. And the world is responding, adopting AI sooner than ever earlier than. What all this progress means is that we’re in a brand new section of the AI platform shift. The place many years of analysis at the moment are changing into actuality for folks, companies and communities all around the world,” Google CEO Sundar Pichai stated in a press release.
The Alphabet-owned firm can be launching a brand new $249.99/month AI subscription plan referred to as Google AI Extremely within the US, providing larger utilization limits and entry to its top-tier AI fashions and options. The brand new plan will probably be rolled out in additional nations quickly, it stated.
Right here is the whole rundown of all of the bulletins from the opening keynote at Google I/0 2025:
Gemini 2.5 Flash and Professional
At Tuesday’s occasion, Google stated it was bringing new capabilities to its most superior AI fashions Gemini 2.5 Flash and Professional equivalent to Deep Assume, an enhanced reasoning mode the place the AI mannequin considers a number of hypotheses earlier than responding to the consumer.
Story continues under this advert
Google stated will probably be making the Deep Assume function out there to a restricted variety of ‘trusted testers’ by way of the Gemini API to get their suggestions earlier than making it broadly out there. Additionally it is integrating 2.5 Flash and Professional with native audio output, superior safety safeguards, and laptop use capabilities.
For builders, 2.5 Professional and Flash will now embody thought summaries and considering budgets within the Gemini API and in Vertex AI. It additionally introduced help for Mannequin Context Protocol (MCP) within the Gemini API for simpler integration with open-source instruments.
Stream, Google’s new AI filmmaking device
Google’s new AI filmmaking device Stream is powered by its most superior AI fashions — Veo, Imagen, and Gemini. It’s able to producing cinematic clips and scenes for films.
Whereas the AI device depends on the Veo and Imagen fashions for producing cinematic visuals, the Gemini integration is supposed to make prompting extra intuitive in order that filmmakers can describe their imaginative and prescient in on a regular basis language.
Story continues under this advert
With Stream, customers can edit and prolong current pictures, management digital camera angles, handle prompts and visible belongings, and entry AI-generated, production-ready clips and content material. Initially often called VideoFX, Stream is now out there to Google AI Professional and Google AI Extremely subscribers within the US, with extra nations coming quickly.
Imagen 4, Veo 3, and Lyria 2
Veo 3 is Google’s newest iteration of its text-to-video generator. It might now generate clips with audio equivalent to site visitors noises within the background of a metropolis road scene, birds singing in a park, and dialogue between characters. The AI device is on the market for Google Extremely subscribers within the US by way of the Gemini app and in Stream. It’s additionally out there for enterprise customers on Vertex AI.
Google has additionally upgraded Veo 2 with new capabilities equivalent to digital camera controls, object add and take away, outpainting, and the power so as to add reference photos of characters, scenes, objects, and so forth.
In the meantime, Imagen 4 has been upgraded to generate 2k decision photos with readability in wonderful particulars like intricate materials, water droplets, and animal fur. The AI-generated photos at the moment are out there to obtain in a spread of side ratios. Imagen 4 can be considerably higher at spelling and typography, Google claimed. It’s out there at the moment within the Gemini app, Whisk, Vertex AI and throughout Gemini Workspace.
Story continues under this advert
Google additional introduced expanded entry to its newest AI music generator Lyria 2, which is now out there for creators by YouTube Shorts and enterprises in Vertex AI. Lyria RealTime, the underlying AI mannequin of MusicFX DJ, can be out there by way of an API and in AI Studio.
Jules, Google’s new AI coding agent
Jules, an AI agent powered by Gemini 2.5 Professional and able to autonomously studying and producing code, made its debut at Google I/0 2025. Customers can combine Jules instantly into their current code repositories. The AI coding agent then makes a clone of the consumer’s codebase inside Google Cloud digital machine (VM) to ‘perceive’ the context of the undertaking and carry out duties equivalent to writing checks, constructing new options, fixing bugs, and so forth.
“Jules operates asynchronously, permitting you to deal with different duties whereas it really works within the background. Upon completion, it presents its plan, reasoning and a diff of the adjustments made,” Google stated. On its safety features, Google stated that Jules is personal by default and doesn’t prepare on a consumer’s personal code.
It’s out there in public beta for everybody, together with Gemini 2.5 Professional free tier customers.
Story continues under this advert
Gemini will get smarter
There have been a number of necessary updates introduced to Google’s Gemini AI assistant on Tuesday. For starters, anybody with an Android or iOS system may have entry to Gemini Dwell’s digital camera and display screen sharing capabilities. Because of this customers can discuss dwell with Gemini about what they see both on their telephone display screen or by the telephone’s digital camera.
Secondly, Gemini is coming to the desktop model of Google Chrome however just for Google AI Professional and Google AI Extremely subscribers within the US. This permits customers to ask Gemini to make clear or summarise any data on any webpage that customers are studying. “Sooner or later, Gemini will be capable to work throughout a number of tabs and navigate web sites in your behalf,” Google stated.