Megadose AI progress, ranked and analyzed.

WorldClaw: Agentic 3D Open-World Generation at Scale

· HF Daily Papers ·
WorldClaw turns open-ended text prompts into large, editable 3D worlds through a coarse-to-fine agent framework.

The paper says planning agents convert a prompt into structured regions, terrain, assets, materials, and spatial relations. WorldClaw then builds a coherent terrain base and fills detail-heavy regions with terrain-conditioned compositions and editable textured meshes. Render-based agents refine terrain, objects, appearance, and contact points. HF Daily Papers' note

score 5

Categories: Research