Roblox has just announced it is releasing an open-source AI model, known as Cube 3D, that is capable of generating 3D objects. The model is one of several initiatives Roblox is undertaking to strengthen the creative capabilities of its platform by utilizing generative AI. The Cube 3D model is powered by an 1.8 billion-parameter AI system, which allows developers and players to produce complex 3D objects with simple text inputs. They can, for example, input a command like “draw a car with racing stripes and a spoiler” and get a corresponding 3D model.
Release of Cube 3D is one of the highlights of Roblox’s vision for streamlining 3D creation. With the release of this model as open-source, Roblox is betting on allowing its community, along with developers external to the platform, to adopt and use the technology. This would accelerate the process of 3D content creation and hopefully attract new types of developers into the platform while increasing the volume of content away from traditional game development.
Roblox’s more long-term ambitions are to push the possibilities of Cube 3D even further with the ability to create whole scenes on demand. This could enable developers to change game environments dynamically, such as the leaves on trees depending on changing seasons. Roblox will also implement other AI-based tools, including text generation, text-to-speech, and speech-to-text functionality, which will be available in the coming months. These are used to enhance the interactive features of games so that more sophisticated NPC interactions become a possibility.
The training data for Cube 3D is a mix of licensed and open-source datasets along with experiential data from the Roblox platform. The blended data allows the model to tokenize 3D objects and read shapes as tokens, ultimately allowing it to forecast upcoming tokens to create a full 3D object. Upcoming releases will see Cube 3D turn into a multimodal model that can create objects from various input sources, including images and videos.
Roblox’s move into AI-driven content creation is just part of its larger strategy to capture a larger percentage of the gaming market. By enabling creators to create new experiences, Roblox hopes to engage more developers and expand its user base. While the generative AI features will be free to developers and creators, usage limits will be implemented to make it viable.
Cube 3D coincides with the time when Roblox is attending the Game Developers Conference (GDC) where the company is showcasing its latest breakthroughs in AI technology. This is in line with Roblox’s emphasis on innovation and that of a leading company in using AI in game development.
Lastly, Roblox’s vision extends beyond the creation of 3D objects to that of 4D generative AI, where dynamic scene creation and changing interactions over time are facilitated. This is an ambitious vision that requires sophisticated knowledge of not just visual look but also physics and scripting to transform the way immersive experiences are designed on the platform.
