Gemini 2.0

Studio Dumbar2023ProjectInterview

14 Sept

Project

Gemini 2.0

Creating a visual language for Gemini 2.0 and the agentic era.

Galaxies of information, a suite of hero films, a catalogue of capability explainers, an entire studio team, an international launch, continuous creative dialogue, 200 icon iterations, moments of connection, AI agents that push the boundaries of technology.

Published
Sept 2025

Agency
Art&Graft

Ahead of its launch in December 2024, Google DeepMind tasked Art&Graft with developing the visual language for Gemini 2.0: Google’s most capable AI model yet, built for the agentic era.

The brief called for our studio team to plug into the DeepMind team’s expertise to fully understand Gemini 2.0’s potential, forming partnerships across design, engineering and brand. This enabled the creation of a conceptual but clear visual system grounded in underlying functionality.

Scale & Rollout

The final deliverables included 9 hero films, 13 motion capabilities, and 49 icons, all encapsulated in a 166 page style guide. An entire website takeover and global launch delivered worldwide impact and comprehensive style guides equip Google’s teams to expand on the visual language and tell new stories.

Development

The visual language used to represent agentic capabilities is characterized by a delicate and ethereal aesthetic, balancing simplicity with complexity. A unified colour palette and foundational structure of lines and dots provide a consistent visual framework.

Subtle gradients, field blurs and orbiting circles set against a dark background present the digital landscape as a galaxy
for the user and agent to navigate through.

The project grew organically and accelerated at speed, building on a successful partnership between Art&Graft
and the Google DeepMind team.

The diverse skillset of our entire studio team was utilised: creatives working across multiple disciplines fed into concepts which were then realised to the highest standard by 3D and motion specialists.

Each asset was thoroughly interrogated: from styleframes,
to icons, to gradient points and motion curves. The result was a coherent visual language for the Gemini 2.0 launch which made complex concepts easily understandable whilst looking beautiful.

Gemini 2.0 Capabilities & Icons

Icons are designed to quickly and effectively convey meaning throughout the website without relying on text, making complex information easier to understand. Built on a modular grid system, any animated icon can transition to another using a simple cut.

Project Astra

Project Astra is a research prototype exploring future capabilities of a universal AI assistant. Built with Gemini 2.0, Project Astra combines advanced real-time conversation, memory, and tool use capabilities.

A galaxy of constellations and pathways formed by dots and lines guide the user through the space, inspired by the experimental nature of a prototype, ancient Greek etymology, and blueprints. Moments of connection are highlighted by flashing glows when the dots collide; embodying the innovation and interaction between the user and the agent.

Project Mariner

Built with Gemini 2.0, Project Mariner is a research prototype exploring future capabilities of human-agent interaction, starting with your browser.

The visual language is characterised by clear, gridded organisational structures and layered frames to represent screens. The agent is represented as a bright yellow dot, which can be seen interacting with the grid when it has identified the most efficient path to complete the user’s task.

Gemini 2.0 For Games

Agents built using Gemini 2.0 can help navigate the virtual world of video games: providing reasoning based on the on-screen action, and offering up suggestions for what to do next in real time conversation.

With a sense of personality, the visual language is characterised by free-flowing interactions and playful curves to reflect the sense of adventure that comes through gaming.