Google has unveiled new features for its Gemini app, including the Gemini Spark AI agent and the Gemini Omni video model. Gemini now has more than 900 million monthly users across 230 countries and in more than 70 languages.
The update also includes a redesign called Neural Expressive, a new Daily Brief agent and a macOS desktop app that will gain Spark integration and new voice tools later.
The broadest immediate change is a redesigned Gemini interface across web, Android and iOS. The new look adds fluid animations, new typography and haptic feedback, while folding the Gemini Live conversational mode into the main app so users can move between typed and spoken interactions.
The microphone experience has also been reworked to let people speak through longer ideas without interruption. Regional dialect options for voices are planned as well.
Gemini is also changing how it presents responses. Rather than returning only text, the app will now generate richer formats in real time, including imagery, timelines, narrated video and graphics.
Video tools
For paying subscribers, Google is adding Gemini Omni, a model that takes text, image and video prompts and turns them into edited video output. It supports tasks such as changing backgrounds, adding cinematic zoom effects and applying templates to footage from a user's camera roll.
Users will also be able to create an AI avatar based on their own appearance and voice. Gemini Omni is rolling out worldwide to Google AI Plus, Pro and Ultra subscribers.
Daily planning
Google is also launching Daily Brief, a morning digest agent that works across connected apps. The feature draws on Gmail and Calendar data to surface urgent updates, upcoming events and follow-up details in a single briefing.
Daily Brief is designed to prioritise information around a user's goals and suggest next steps. It is being introduced to Google AI Plus, Pro and Ultra subscribers in the US.
Agent shift
The most significant addition is Gemini Spark, a 24-hour personal AI agent that can manage tasks in the background. Built on Gemini 3.5 and integrated with Workspace products such as Gmail, Docs and Slides, Spark is intended to move Gemini beyond answering prompts and towards carrying out multi-step work under user control.
Google says Spark can review monthly credit card statements to flag subscription fees, monitor school-related emails for deadlines, and turn meeting notes from emails and chats into Google Docs and draft project emails.
Because Spark is cloud-based, it can continue operating when a laptop is closed or a phone is locked. Users choose whether to enable the agent and which apps it can access, and it is designed to ask for approval before actions such as spending money or sending emails.
Spark is being released first to trusted testers, with a beta launch planned for US Google AI Ultra subscribers. Google is also extending Gemini's connections to third-party services including Canva, OpenTable and Instacart.
Mac desktop
Google has also released a Gemini app for macOS. Later updates will allow Spark to work with local files and automate workflows on the machine itself.
Voice features are also coming to the Mac app, using screen context to turn spoken input into written drafts at the cursor location. The system is intended to handle disfluencies in natural speech and convert them into cleaner text.
The announcement comes as Google pushes to make Gemini a more central consumer AI product, with a greater focus on persistent agents and multimodal creation. Growth from 400 million to more than 900 million monthly users suggests the company is broadening its reach as competition intensifies across consumer AI assistants.