Roblox Launches Cube 3D and Mesh Generation API for AI-Powered 3D Design

In General XR News

March 19, 2025 – Roblox, the global immersive platform for creation, communication, and connection, has this week announced the open-source release of Cube 3D, a 3D foundational model that allows for the creation of 3D objects and environments on the Roblox platform directly from a text-based prompt. 

As part of the release, Roblox has launched the first of Cube 3D’s capabilities, with the beta launch of the company’s Mesh Generation API, enabling developers to create 3D assets more efficiently.

Roblox stated that Cube will underpin many of the AI tools the company plans to develop in the years to come, including highly complex scene-generation tools. Ultimately, Cube will be a multimodal model, trained on text, images, video, and other types of input, and will integrate with Roblox’s existing AI creation tools. Future versions of Cube will also be able to generate 3D models and environments from image inputs, according to the company.

Cube 3D enables rapid generation of 3D objects.

AI-Driven 3D Creation

Cube 3D is trained on native 3D data from the Roblox platform, distinguishing it from existing methods that rely on 2D image reconstruction. The model has been built to understand shapes as tokens, and essentially tokenizes 3D objects in a manner similar to how large language models (LLMs) process words.

As a result, Cube 3D has been trained to predict the next shape in a sequence, allowing it to build complete 3D objects. Extending this ability to full scene generation, Cube 3D predicts the layout and recursively predicts the shape to complete that layout. Furthermore, anyone can fine-tune, develop plug-ins for, or train Cube 3D on their own data to suit their needs, according to Roblox.

“Players can personalize their experience in ways developers never imagined, and that’s going to make their games even more engaging.”

Empowering Developers with AI Tools

Over the past year, Roblox has integrated AI-powered tools into its Roblox Studio development platform, including an AI Assistant designed to streamline 3D creation. With the introduction of Cube 3D and the Mesh Generation API, developers can generate assets more efficiently, reducing manual work and allowing for quicker iteration on design ideas.

The beta launch of the Mesh Generation API allows developers to generate 3D assets in a matter of seconds by entering simple prompts. These generated objects are compatible with game engines and can be further refined with textures, colors, and other details. Roblox expects that AI-driven tools like Cube 3D will significantly reduce the time required for asset creation, allowing developers to focus on broader design aspects rather than spending hours on individual models. Longer term, Roblox plans to enable more complex and functional objects, even scenes.

A couch asset created using the prompt: “A vintage green couch with clean lines and velvet material.”

In a blog post announcing the launch of Cube 3D, Anupam Singh, Vice President of Engineering, and Nick Tornow, Vice President of Creator Engineering at Roblox, said: “We see a future where developers enable their users to become creators using AI. With the Mesh Generation API enabled, players can bring to life anything they can imagine… Players can personalize their experience in ways developers never imagined, and that’s going to make their games even more engaging.”

Open-Source AI and Future Development

Roblox stated that in order to address the key technical challenge of connecting text and images with 3D shapes, it developed a unified architecture for 3D generation using autoregressive transformers, a type of neural network that predicts the next component based on previous inputs.

According to the company, this approach enables both scalability and multimodal compatibility, allowing for the processing of text, visual, audio, and 3D inputs as the model expands. Roblox has open-sourced its model, and for this initial stage, creators will be able to generate 3D objects based on text prompts.

Cube 3D’s architecture processes text and image inputs using ShapeGPT for 3D shape generation and LayoutGPT for spatial organization, enabling structured scene creation from text prompts.

 

What Will Cube Be Able to Do in the Future?

Beyond individual object creation, Roblox plans to expand Cube 3D’s capabilities to full scene generation. The company stated that it envisions AI assisting with layout design and object placement, enabling a more interactive and dynamic virtual environment where different components are functional within a scene and within the context of one another. Roblox refers to this as “4D creation,” where the fourth dimension is interaction between objects, environments, and people.

In the long term, Roblox sees Cube enabling in-game AI generation, allowing players to create and personalize assets dynamically. Future updates are expected to further integrate Cube into the platform’s AI ecosystem.

The open-source version of Cube 3D will be available later this week.

Image / video credit: Roblox

About the author

Sam is the Founder and Managing Editor of Auganix. With a background in research and report writing, he has been covering XR industry news for the past seven years.

Sign Up to the Weekly Auganix Newsletter