Features
Generate 3D world from a single image
Generate 3D world from video input
Generate 3D world from text prompt
Generate 3D world from 3D layout
High-fidelity persistent geometry
Realistic lighting and materials inference
Editable and exportable 3D scenes
Fast generation in seconds
Browser-based interface, no local installation required
Supports complex indoor and outdoor scenes
Consistent world across multiple viewpoints
Collaboration sharing features
Generates textured 3D meshes from 2D image collections
Supports complex topology and detailed geometry
End-to-end training with only 2D supervision
Differentiable rendering for high-fidelity textures
Multi-category generation (cars, chairs, animals)
Outputs explicit mesh and texture maps
Open-source code on GitHub
Works with standard graphics pipelines
