Building the largest reality-grade 3D dataset
We're building a dataset that redefines what’s possible in 3D content generation — ultra-realistic, relightable, and at massive scale. Already used by Meta Reality Labs, now available to the wider research and enterprise ecosystem.
A new standard for 3D datasets
While generative AI continues to advance, most available training datasets remain limited — and crucially, not 3D-native. Large-scale collections like Objaverse, Google Scanned Objects, and ShapeNet provide volume but fall short on realism, consistency, and physical accuracy.
ALLSIDES bridges this gap with reality-grade, relightable 3D data purpose-built for machine learning applications. Capturing real-world objects at industrial scale, our dataset doesn’t approximate reality — it preserves it.
Experience it yourself. Download a selection of raw scans, straight from our capture pipeline.
The Digital Twin Catalog by Meta
We partnered with Meta to release the Digital Twin Catalog, the most photorealistic open-source 3D dataset to date. Our scanners powered the object capture pipeline behind the dataset.
Comparison Table
Existing Solutions | Our Solution | |
| QUALITY | Synthetic data or low-fidelity assets. | High quality real-world objects: sub-mm geometric accuracy, HD pbr textures. Physically accurate. |
| LIGHT | Lighting is often baked into textures. | All assets are relightable — enabling fully controllable rendering in any scene or lighting condition. |
| GEOMETRY | Low poly with poor surface details and broken topology. | Clean and watertight meshes, with high-fidelity surface. |
| METADATA | Inconsistent metadata. | Objects are tagged with structured metadata (category, size, reference images, physical properties, etc.). |
| DIVERSITY | Limited variety across real world object categories, textures, shapes... | Real consumer goods (food, natural items, accessories, kitchenware...) spanning different shapes, materials, and use cases. |
At a glance
A digital archive of physical objects. Our vast dataset spans countless categories and materials, creating the most comprehensive 3D model repository for advancing computer vision, Al research, content creation and immersive digital experiences.
+ Multiple formats: GLB, USDZ, OBJ.
+ Structured metadata per asset (material, category, dimensions...).
Want access to the dataset?
We’re actively onboarding select research and enterprise partners. Reach out to discuss licensing, collaboration, or use cases.