A23D — The largest proprietary library of human-authored 3D assets — built for AI training and generative 3D

A23D is a proprietary dataset of 500,000+ 3D models, PBR textures, decals, and IES profiles, entirely human-authored in-house. Every asset is built by our own team to one concept, one vision, one structure — delivering consistency across scale, orientation, topology, materials, and metadata that open-source and marketplace datasets can't match.

What Makes It Different

Every model is textured with our own PBR materials, giving us component-level geometry-and-material ground truth (albedo, roughness, metallic, normals). Every asset in the corpus contains linked geometry, materials, topology, taxonomy, dimensions, and component-level metadata created under a single production standard.

This forms a structured graph connecting geometry, materials, semantics, topology, dimensions, and categories — a depth of per-asset metadata not found in open-source or marketplace datasets.

Dataset at a Glance

• 400,000+ 3D Models
• 100,000+ PBR Textures
• 14,000+ Decals
• 10,000+ IES Profiles

Use Cases

• Training data for generative 3D and 3D foundation models
• Geometry-and-material ground truth for reconstruction and material estimation
• Structured 3D corpora for research in topology, semantics, and taxonomy

Licensing

We license this dataset for AI training and generative 3D. To discuss access, reach out via enterprise@a23d.co or visit a23d.co/ai