Spatial Edit (Apache 2.0)

Developer Eason Xiao has launched SpatialEdit-16B, a significant open-source contribution to the AI image editing space. The 16-billion parameter model, released under the permissive Apache 2.0 license on GitHub and Hugging Face, enables 3D-aware manipulation of objects within images using natural language prompts. Unlike traditional 2D editing tools, SpatialEdit understands spatial relationships between objects, allowing users to move items forward/backward, change perspectives, or adjust lighting in a coherent 3D context.

What sets SpatialEdit apart is its ability to interpret spatial commands like "move the chair to the left" or "make the building appear farther away" while maintaining realistic lighting, shadows, and object interactions. The model demonstrates particular strength in handling complex scenes with multiple objects, preserving the overall scene coherence during edits. Early adopters report successful applications in product photography adjustments, architectural visualization tweaks, and creative scene composition without needing 3D modeling expertise.

The open-source nature means developers can integrate SpatialEdit's capabilities into existing workflows or build custom applications. While currently at 16B parameters, the model shows promising results for a mid-sized architecture, balancing computational requirements with editing quality. As 3D-aware AI editing becomes more accessible, it could democratize professional-grade image manipulation previously requiring expensive software and specialized skills.