The pioneering AI 3D research company unveils its most advanced model to date, a 10B-parameter 3D generative AI built on the BANG architecture, geared for professional uses across film, TV and games.
Deemos has unveiled Rodin Gen-2, its most advanced 3D generative AI model to date.Built upon the groundbreaking BANG architecture, which bridges 3D generation and reasoning, and trained at an unprecedented scale — reaching 10 billion parameters — Rodin Gen-2 sets a new benchmark for quality and control in AI-powered 3D creation.
Rodin Gen-2 introduces a paradigm of recursive part-based generation, allowing it to intelligently divide and subdivide objects into coherent components.This enables the creation of highly complex and structured 3D assets directly from textual or visual prompts.Compared with earlier models, Rodin Gen-2 delivers a 4x improvement in geometric mesh quality, expertly balancing intricate surface details with structural regularity.
Furthermore, it supports baked normals and HD texture maps (Beta), allowing creators to display high-polygon details on optimized, low-polygon models — a critical feature for real-time applications in game development.Ready to experience this generative breakthrough? Visit hyper3d.ai now to register for your complimentary 14-day Creator Plan.Use the code awnRodinGen2 during sign-up for instant access.
Backed by Award-Winning Research Rodin Gen-2 builds on Deemos’ world-class research track record.In just four years, the team has published 16 top-tier papers and delivered breakthroughs recognized across the field, including: CLAY — a Controllable Large-scale Generative Model for Creating High-quality 3D assets, awarded Best Paper Honorable Mention at SIGGRAPH 2024.CLAY has since become a mainstream framework for 3D generation CAST — a physics- and occlusion-aware scene reconstruction method, awarded Best Paper at SIGGRAPH 2025 BANG — a paradigm-shifting framework for asset decomposition, published in ACM Transactions on Graphics and named among Top 10 Technical Papers Fast Forward at SIGGRAPH 2025 These peer-reviewed honors underscore the scientific foundation behind Rodin Gen-2 and its potential to transform both research and production pipelines.
Precision Meets Practicality While Rodin Gen-2’s breakthrough lies in its new BANG architecture, it also continues to support 3D ControlNet, a widely adopted industry technique that Deemos first integrated in earlier versions of Rodin.With bounding box, voxel, and point cloud control available, creators can guide generation with precise spatial input, ensuring Rodin Gen-2 is not only powerful but also highly practical for professional workflows.A New Paradigm for 3D Generative AI With its new BANG architecture, massive scale, and integrated controllability, Rodin Gen-2 doesn't just generate shapes.
It understands and constructs objects recursively from parts, resonating with how creators think.This is a foundational shift towards true component-aware design and manufacturing.The model's core "Generative Exploded Dynamics" technology provides a unique window into its reasoning, allowing users to visualize how generated objects decompose into their constituent parts.
This offers unparalleled control, extendable through spatial prompts and multimodal models, for intuitive editing and manipulation.Rodin Gen-2's capabilities open up a wide range of professional applications, from accelerating high-quality game asset creation to facilitating 3D printing workflows by generating pre-separated, easy-to-print parts.Unlock your creative potential with Rodin Gen-2.
Don't forget to use your code awnRodinGen2 for a free 14-day trial of the Creator Plan at hyper3d.ai.Follow Deemos on X (Twitter) at @DeemosTech for the latest updates.About Deemos Deemos is a pioneering AI 3D research company dedicated to making 3D creation as intuitive as imagination itself.
We build powerful 3D generative AI models that bridge the gap between creative vision and 3D reality, empowering creators across industries.Founded by leading experts in AI and computer graphics, we are defining the future of intelligent 3D content creation.Dan Sarto is Publisher and Editor-in-Chief of Animation World Network.
Jeremy Renner Boards ‘Stardust Future: Stars and Scars’ AI-Animated Feature
FOX Entertainment Invests in Holywater
NVIDIA Open Sources Audio2Face Animation Model
Previs Pro Gets Major Update, Free Access Tier
OPSIS Partners with Utopai Studios to Pioneer AI-Supported Content Creation
Outsyders Reveals AI-Powered Immersive Projects for Disney, IMAX, Samsung/Google XR
Lightcraft Unveils ‘Spark’ AI-Powered Production Platform for Filmmakers
