Roblox, the metaverse platform, has launched two new generative AI tools in beta for game developers, known as Code Assist and Material Generator. These features are part of Roblox’s generative AI strategy, which aims to accelerate and streamline object design and deployment within its digital universe.
Code Assist is designed to automate basic coding tasks, allowing developers to focus on their creative work, said Roblox Studio head, Stef Corazza, at the Game Developer’s Conference. With Code Assist, users can translate natural language into code and embed it into a metaverse space or object. The tool doesn’t work without other code to call on, but it can help with filling in the details or more repetitive aspects of coding.
In his speech, Corazza demonstrated Code Assist’s capabilities, changing the colour of an orb to red, destroying it 0.3 seconds after a player touched it, and surrounding the orb with a grid of other orbs. The tool then wrote code to add new orbs and adjust how it reacted to player interaction.
The Material Generator tool produces art assets that can be layered on two-dimensional surfaces, simulating different maps of roughness, albedo, and “metalness.” The game engine can then reference those factors to ensure that the object looks and behaves appropriately. The aim is to produce 3D assets for an entire game based on a few concept drawings.
The long-term goal is to enable developers to create interactive 3D worlds by typing simple instructions, such as “scene with a forest, a river, and a large rock,” into an interface. Roblox also plans to encourage creators to embed all kinds of generative AI tools into its metaverse.
Roblox’s generative AI strategy combines in-house and unnamed licensed models, with training restricted to public-domain game content. Recent advances in generative AI models have made the development of these tools quicker, but Corazza cautioned that they are not perfect and should not be blindly incorporated into a game. However, if an error is detected, it is easy to redo the work.
“It is still up to you to review, test, and determine if the code suggestion is contextually appropriate,” said Corazza. “No catastrophic events will happen if the generation isn’t good—just click the button and create another one.”
Roblox creators are already using text-to-image AI tools such as DALL-E and Midjourney, and OpenAI’s POINT-E text-to-object generative AI tool has opened up new possibilities. Microsoft’s GitHub Copilot AI coding assistant, which uses OpenAI’s language models, has also become very popular.
Roblox’s generative AI tools are designed to help those who are blocked by technical hurdles but have great ideas. With these new features, Roblox is well on its way to making creating within the metaverse accessible to everyone. The possibilities for creating interactive 3D worlds are endless, and Roblox plans to continue to encourage the development of new generative AI tools.