)
Overview
When someform began developing the DeepMind 3D Design System, it was clear from the outset that the ultimate goal would be the creation of a standalone application tailored to the needs of the design team at Google DeepMind.
This application would empower the in-house team to effortlessly craft images within a cohesive visual language.
someform’s conceptual ideation therefore aimed at achieving a dual objective: on one hand, to forge a distinctive and recognizable visual language that embodies the core principles of DeepMind; on the other, to ensure that this design system would serve as the foundation for an independent and versatile creative application. With this overarching objective in mind, someform embarked on the initial stages of crafting a 3D toolkit—one designed to possess uniquely ownable visual characteristics while providing the flexibility needed to facilitate a wide array of compositions and narratives.
Because each element within the system draws inspiration from specific aspects of Artificial General Intelligence (AGI), someform’s commitment extended beyond creating aesthetically compelling visuals. It was essential to underscore the narrative behind each abstraction, ensuring a harmonious blend of design and conceptual depth.
Following the completion of the project’s first phase—which delivered the ruleset of a comprehensive 3D design system and a collection of traditionally rendered assets—someform continued into the development of the standalone application.
This application is designed to leverage the system, enabling users to craft new images within a consistent visual language. Beginning with user experience and user interface, someform’s paramount goal was to create an app that feels intuitive and approachable.
The DeepMind Image Tool is an intuitive standalone application built on Unreal Engine 5 that allows users to create complex imagery using a progressive, hardware-accelerated path tracer. All images on this page were created using the DeepMind Image Tool.
:quality(50))
The user experience unfolds through a carefully sequenced series of steps and chapters—a deliberate design choice that avoids the common pitfall of overwhelming users with excessive options. The application strikes a balance between user-directed interaction and semi-procedural algorithms, streamlining many of the tedious tasks required to achieve a specific visual language. This blend ensures the software is not only user-friendly but also clarifies its functionality as users navigate, ultimately providing a robust system for generating distinctive compositions and layouts.
Recognizing that not every user would have prior experience with 3D software, someform prioritized designing an interface that serves both novices and experienced users alike.
)
:quality(50))
【chapter 1: blocking】
The first steps emphasize initial spatial layout considerations, postponing the creation of specific frames until later in the process. This approach ensures that users can easily navigate the 3D layout of their scenes before refining frame-specific details.
The scene layout is constructed through the placement of simple cubes, which users can position and move within a defined volume. These cubes automatically align to a spatial grid, ensuring precision and consistency.
Once the foundational layout is established, the application applies a suite of procedural algorithms to transform these basic shapes into what are referred to as “Agent shapes.”
)
)
【chapter 2: editing】
The second chapter shifts attention to the secondary elements within the scene: the “Model shapes” and the generation of connecting lines. After placing the initial subtractive model shapes earlier in the process, users can edit, create, or delete these elements in this phase.
To maintain visual consistency, someform sought a balance between customizability and constraint. For example, users may scale objects—but within predefined limits that ensure adherence to the design system’s parameters. This approach preserves creative freedom while ensuring coherency within the visual aesthetic.
:quality(50))
)
【chapter 3: rendering】
The final chapter focuses on creating the finished frames. At this stage, the application grants full freedom to navigate the 3D environment, offering a suite of tools to adjust camera position and rotation.
The intention was to evoke the immersive experience of a photographer working within a defined scene. The goal is to emulate a photographer’s pursuit of the ideal framing, offering tools to refine camera settings and lighting so users can concentrate purely on composition.
Once users achieve their desired frame, the application allows them to render high-quality images up to 8K resolution within minutes. And with its strong focus on data, the tool enables users to save projects, revisit scenes at any time, adjust models or layouts, reposition the camera, and generate new frames from previously established compositions.
)
)
)
)
【conclusion】
At the project’s inception, someform aimed to forge a distinctive and recognizable 3D design system tailored for Google DeepMind. Once established, this versatile system transcended its initial purpose, finding expression across multiple formats—from traditionally rendered images to intricate animations—and ultimately culminating in the creation of a standalone application.
This application now empowers the DeepMind design team to harness their custom visual language, enabling them to craft unique narratives within the framework of the 3D design system.
)
:quality(50))
:quality(50))
:quality(50))
:quality(50))
:quality(50))
:quality(50))
:quality(50))
:quality(50))
:quality(50))
:quality(50))