EZBlender: Efficient 3D Editing with Plan-and-ReAct Agent

Hao Wang, Wenhui Zhu, Shao Tang, Zhipeng Wang, Xuanzhao Dong, Xin Li, Xiwen Chen, Ashish Bastola, Xinhao Huang, Yalin Wang, Abolfazl Razi; Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops, 2026, pp. 1343-1352

As a cornerstone of the modern digital economy, 3D modeling and rendering demand substantial resources and manual effort when scene editing is performed in the traditional manner. Despite recent progress in VLM-based agents for 3D editing, the fundamental issue between editing precision and agent responsiveness remains unresolved. To overcome these limitations, we present EZBlender, a Blender agent with a hybrid framework that combines planning-based task decomposition and reactive local autonomy for efficient human-AI collaboration and semantically faithful 3D editing. Specifically, this unexplored Plan-and-ReAct design not only preserves scene-editing performance but also significantly reduces latency and computational cost. Compared with state-of-the-art 3D editing agents, EZBlender achieves up to a seven-fold improvement in response speed and reduces token consumption by 67%. Beyond these quantitative gains, the framework also demonstrates strong text-visual prompt alignment, a capability crucial for high-quality scene editing. To further validate the efficiency and effectiveness of the proposed edge-autonomy architecture, we construct a dedicated multi-tasking benchmark that has not been systematically investigated in prior research. In addition, we provide a comprehensive analysis of language model preference, system responsiveness, and economic efficiency.