SpaceFlow: Locally Controllable 3D Generation
SpaceFlow enables locally controllable, training-free 3D generation from text and geometric primitives.
SpaceFlow is a training-free pipeline for locally controllable 3D generation from text and geometric primitives. Each primitive proxies an object part and carries a local control level, so high-control regions follow the input shape while low-control regions allow generative completion. Appearance is synthesized per segmented part from its assigned text or image cue, limiting cross-part leakage. Regional metrics and a user study show competitive geometric fidelity, and text-conditioned routing is state of the art for prompt faithfulness and color or material accuracy on fixed geometry.
- Training-free pipeline assigns a local control level to each geometric primitive.
- Structure generation enforces spatial constraints inside the generative flow.
- Part-wise appearance conditioning limits cross-part leakage.
- Text routing leads prompt faithfulness and material accuracy on fixed geometry.
From the original article
1 min read at huggingface.co
Current 3D generation methods lack explicit local control: geometric adherence is often defined by a global control strength, and appearance cannot be specified locally. We present SpaceFlow, a training-free pipeline for locally controllable 3D generation from text descriptions and a collection of geometric primitives.
By Neil De La Fuente, Joan Lafuente, Mukhammadali Sayfiddinov · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.