Roblox, a popular gaming platform, has introduced its open-source Cube 3D AI model. This is the company’s first iteration of its foundational AI model capable of generating 3D objects. The company said that the beta version of Cube 3D mesh generation will be available this week in Roblox Studio and as an in-experience Lua API.
“Cube will underpin many of the AI tools we’ll develop in the years to come, including highly complex scene-generation tools. It will ultimately be a multimodal model, trained on text, images, video, and other types of input—and will integrate with our existing AI creation tools,” the company said in a statement.
What is Cube 3D?
Cube 3D is a foundational AI model that generates 3D models and environments directly from text. It will likely generate output from image inputs in the future. Unlike conventional methods that heavily rely on 2D image reconstruction to build 3D models, the company claims that Cube 3D is trained on native 3D data, which makes its outputs functional and immersive.
The company is outsourcing a version of Cube 3D, essentially allowing developers and researchers to fine-tune, customise, and integrate the model into their applications. Among the first capabilities, the company has introduced the Mesh Generation API, which is available in Roblox Studio and Lua API, allowing developers to create 3D objects from simple text prompts.
How does Cube 3D work?
The company said that Cube 3D uses an advanced AI technique known as 3D tokenisation, which converts 3D shapes into tokens. This is similar to how language models break text into words. Reportedly, this allows AI to forecast and generate the next shape token to build complete 3D objects. Over time, this approach will lead to the generation of full scenes. Cube 3D is based on autoregressive transformers, a type of transformer model that is often used for text generation. Since the model has a scalable architecture, the AI model can handle various types of inputs, including text, images, and video, making it a multimodal AI model.
At present, Cube 3D lets users generate 3D assets like furniture, vehicles, clothing, weapons and animals using simple text prompts. The company has said that future capabilities will extend to scene creation, object interactivity, dynamic content generation, etc.
Source: https://indianexpress.com/
