Spatial Intelligence / 3D Generation

Build 3D Worlds with AI

Request Early Access Learn More
0Cost Reduction
0Faster Production
0Learning Curve
// Industry Pain Points

Traditional 3D Creation
Needs a Revolution

3D content creation still relies on expensive software, specialized artists, and weeks of manual labor. The gap between imagination and creation remains enormous.

Prohibitive Costs

Professional 3D software licenses cost tens of thousands of dollars. Creating a single interior scene requires a senior artist working an entire day.

Steep Learning Curve

Maya, 3ds Max, Blender -- each tool requires months of training. Creative visions get blocked by technical barriers.

Severe Talent Shortage

Demand for 3D content in gaming, AR/VR, e-commerce, and the spatial web is exploding, yet qualified professionals remain scarce.

Low Reusability

3D models are typically locked to a single scene, suffer poor cross-platform compatibility, and are difficult to re-edit after creation.

AI-Generated 3D Scene Demo
// Core Capabilities

One Engine, Infinite Possibilities

Stellar Engine unifies spatial intelligence, generative AI, and real-time rendering into a single platform, empowering anyone to create 3D content.

3D Scene Generation

Describe your scene in natural language or upload a reference image. Our spatial intelligence model automatically generates a complete 3D environment with realistic object placement, lighting, and spatial relationships.

3D Scene Generation Example

3D Asset Creation

Generate production-ready 3D models from a single image or text prompt. Outputs clean meshes with PBR materials, compatible with game engines, AR applications, and commercial projects.

AI-Generated 3D Model

Real-Time 3D Reconstruction

Capture any physical space with a standard RGB camera. Our reconstruction engine instantly generates high-fidelity 3D models -- no LiDAR or depth sensors required.

Real-Time 3D Reconstruction Demo
// Technology

Proprietary AI Core Technology

Built on breakthrough research in spatial intelligence and differentiable rendering, Stellar Engine redefines the boundaries of AI-powered 3D creation.

M-DRG 3D Representation

Our proprietary Mesh-based Differentiable Rendering Gaussian (M-DRG) is a novel 3D representation that fuses the geometric precision of meshes with the rendering quality of Gaussian splatting. It enables element-level editing, physics simulation, and real-time rendering in a unified representation.

Geometry-AccurateGaussian RenderingPhysics-ReadyReal-Time InteractiveElement-Independent
M-DRG Representation Comparison

Spatial Intelligence System

Our spatial reasoning module understands 3D spatial relationships from 2D inputs, performing amodal segmentation, depth estimation, and spatial layout prediction to reconstruct complete 3D scenes from limited observations -- powered by a dedicated Vision Transformer architecture.

ViT EncoderSpatial ReasoningAmodal SegmentationDepth EstimationMultimodal Input
Spatial Intelligence Pipeline

AI-Driven Dynamic Physics

Moving beyond hand-coded physics rules, we use large language models and visual generative models to drive dynamic events in 3D scenes. Objects interact naturally based on spatial context and learned physical priors -- no traditional physics engine required.

LLM-DrivenEvent SystemDiffusion ConstraintsDynamic Scenes
Dynamic Physics Event-Driven Scene
// Workflow

Three Steps from Idea to 3D

No modeling experience required. Describe what you want and let AI handle all the complexity.

01

Input

Provide text descriptions, reference images, sketches, or video clips. Specify style, mood, and layout using natural language.

02

AI Generation

Spatial intelligence decomposes the scene, generates 3D elements one by one, and automatically assembles them with correct spatial relationships.

03

Render & Export

M-DRG delivers real-time rendered output. Export to Unreal, Unity, Blender, and web viewers.

// Showcase

AI-Generated 3D Assets

Production-grade 3D models generated by Stellar Engine. All assets are commercially licensed with clean topology, PBR materials, and standard export formats.

Commercial
Character 3D Model

Character Assets

mesh + pbr textures // .glb .fbx .obj

Commercial
Interior Scene 3D Model

Interior Scene

full scene + lighting // .glb .usd

Commercial
Architecture 3D Model

Architecture Model

exterior + interior // .glb .usd

// Use Cases

3D Solutions Across Every Industry

From indie game studios to enterprise-grade 3D pipelines, Stellar Engine integrates seamlessly into your workflow.

Game Development Scene

Game Development

Batch-generate open-world environments, characters, and props. Export directly to Unreal Engine and Unity.

E-Commerce & Digital Twins

E-Commerce & Digital Twins

Build photorealistic 3D product showcases and digital replicas of physical spaces, empowering real estate, retail, and museums.

// Team

Research-Driven, Product-First

Founded by a research team from the Chinese Academy of Sciences, with deep expertise in 3D reconstruction, computer vision, and artificial intelligence.

Yiming Cai

Yiming Cai

CEO & CTO

Ph.D. in Optical Engineering from Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences. Recipient of the National Science and Technology Progress Award (First Prize). Long-standing researcher in 3D reconstruction, stereo mapping, and AI synthetic data -- one of the earliest AI synthetic data researchers in China.

Fangjie Yu

Head of Product

M.S. from Changchun Institute of Optics, CAS. Former algorithm team lead at Sensing Photonics, where she led multiple B2B solutions from 0-to-1 in R&D and deployment, achieving over 60% market share.

Yawei Hou

Algorithm Engineer

M.S. in Computer Vision. Former algorithm lead at Jingwei Information. Published multiple papers and patents. Previously led an AR glasses algorithm team, developing industry-leading 3D hand reconstruction and interaction algorithms.

Tianjie Ji

Algorithm Engineer

M.S. from Shenyang Institute of Computing Technology, CAS. Emerging researcher in 3D reconstruction, deeply involved in multiple major 3D reconstruction research projects.

Yujing Ren

Engine Engineer

M.S. from University of Leeds, specializing in high-performance game engines. Emerging researcher in amodal segmentation and inpainting diffusion.

Xiaoning Lv

Advisor (Part-time)

M.S. from the Aerospace Information Research Institute, CAS. Renowned computer vision scholar with dozens of published CV papers, with extensive experience in major national science and technology projects.

// Get Started

Ready to Build the Future of 3D?

Join our early access program and create production-grade 3D content with AI. API and platform access available for developers and studios.

Contact Us