Roblox Company, a worldwide platform for immersive user-generated content material, has launched Dice 3D, a foundational generative AI mannequin for creating 3D digital content material from textual content prompts. The corporate has made a model of the mannequin open-source, now accessible through GitHub and HuggingFace. The beta model, which integrates into Roblox Studio and its Lua-based API, permits creators to supply 3D meshes immediately inside recreation environments utilizing pure language enter.
In contrast to image-based reconstruction strategies that depend on restricted visible knowledge, Dice 3D is educated on native 3D belongings generated and used throughout Roblox’s ecosystem. This permits it to output structurally full digital objects appropriate with recreation engines—objects that may be interacted with in gameplay, reasonably than being flat facades. A typical use case includes getting into a command like “generate a motorbike” into the platform’s Assistant, leading to a full 3D mesh appropriate for speedy in-game deployment. These objects can later be enhanced with texture and colour however are generated as functionally usable meshes from the outset.
Dice 3D applies a token-based system to know and predict 3D shapes. Drawing from strategies utilized in giant language fashions, the AI converts geometry into form tokens and makes use of autoregressive transformers to forecast subsequent tokens in a sequence—successfully “constructing” a mesh piece by piece. This technique helps each particular person object completion and full scene structure era. To align multimodal inputs, Roblox engineers developed a unified transformer structure appropriate with textual content, picture, and future knowledge sorts akin to audio. The present launch focuses on object era from textual content, however future updates are anticipated to assist scene-level outputs and hybrid enter modalities.
Roblox has positioned the software program as a part of a broader shift towards real-time, user-augmented content material creation. Gamers and builders alike will be capable to generate props, environments, or interactive objects on demand. The longer-term intention is “4D creation,” the place AI understands not solely object kind but additionally interplay logic and environmental relationships. This contains bounding field placement for structure, mesh fusion for multi-object environments, and context-aware alterations—akin to swapping seasonal parts or adapting geometry primarily based on narrative triggers inside a recreation.
Whereas Dice 3D doesn’t at the moment assist 3D printing file codecs akin to STL, the underlying methodology of tokenizing 3D shapes might affect rising instruments in digital prototyping, AI-assisted design, and even CAD automation. Open-source launch of this nature stays uncommon amongst proprietary recreation growth platforms, significantly these coping with native 3D asset pipelines. Roblox has additionally co-founded ROOST, a nonprofit targeted on open-source AI security, and just lately launched different fashions tied to accountable AI growth.
Rising AI-generated 3D modeling
Tencent, a Chinese language multinational expertise firm with important investments in gaming and cloud providers, launched Hunyuan3D 2.0 to streamline digital asset creation. The system options two specialised fashions—Hunyuan3D-DiT for geometry and Hunyuan3D-Paint for texture—designed to enhance constancy and responsiveness in 3D era. Inner benchmarks utilizing Situation-Mannequin Matching Distance (CMMD) and Fréchet Inception Distance (FID) recommend larger alignment between person prompts and ensuing outputs. A companion interface, Hunyuan3D-Studio, permits sketch-to-3D workflows and low-polygon mesh export. Whereas Tencent has not formally focused additive manufacturing, the platform’s skill to generate each high-resolution and simplified meshes might assist adaptation in prototyping and multi-material 3D printing environments.
A 12 months earlier, Nvidia, a number one developer of GPUs and parallel computing platforms extensively utilized in AI and visualization, launched Magic3D in 2023 as a part of its generative AI analysis. The instrument produces textured 3D meshes from pure language utilizing a two-stage pipeline that refines coarse fashions into high-resolution geometry. Demonstrations—akin to rendering a blue poison dart frog from a single immediate—illustrated the system’s capability for mesh synthesis, text-driven modifying, and stylistic transformation. Though primarily geared towards video video games and computer-generated imagery, Nvidia researchers recognized potential crossover purposes in VR growth and particular results, noting that streamlined mannequin era may decrease manufacturing obstacles throughout digital design domains.
Prepared to find who gained the 2024 3D Printing Trade Awards?
Subscribe to the 3D Printing Trade e-newsletter to remain up to date with the most recent information and insights.
Featured picture showcase Immediate: A purple buggy with knobby tires. Picture through Roblox.
