Video-Coder trains agent to generate videos using 3D scene code
Video-Coder: Reinforcing Agentic Coding for Video Generation
Computer Vision and Pattern Recognition
Summary
Making realistic videos that show correct shapes, smooth motion, and easy control is hard. The researchers created Video-Coder, a system that learns to write and run code building 3D scenes which guide video creation. They trained the system on new video and code data with rewards based on how well the code runs and how good the videos look. This approach improved video generation quality and offers an open framework for future coding agents in video creation.
What this means in practice
- •For video content creators: Generate videos with controllable scene elements and realistic motion by automatically coding dynamic 3D scenes for downstream video synthesis.$Commercial implications: Enables selling tools for creating customizable, high-quality videos for entertainment, advertising, or simulation industries.
- •For game developers: Produce coherent animated video clips with controllable spatial and motion elements to use as game cinematics or environmental assets.
Authors
Chaoyang Wang, Sicheng Gao, Junan Hu, Kunyang Li, Dongyang Chen, Yiyang Zhou, Huaxiu Yao, Yong Jae Lee, Radu Timofte, Yuzhang Shang
Abstract
Recent years have witnessed remarkable progress in video generation in terms of visual quality and temporal consistency. However, generating videos with precise spatial structure, coherent motion dynamics, and controllable scene composition remains challenging. In this paper, we introduce Video-Coder, the first attempt to train a code-driven video generation agent, which autonomously writes and executes code to construct a dynamic 3D scene (Draft Video) that provides structure and motion signals to downstream video generators. To support this paradigm, we design a dedicated data pipeline and curate two high-quality datasets, Video-Coder-SFT-6k and Video-Coder-RL-2k, which contain diverse prompts and ground-truth reference videos. We further develop the Video- Coder Harness, a unified interactive environment integrating code execution, 3D scene rendering, and execution feedback for agent training and inference. We then train Video-Coder with SFT fol- lowed by harness-native agentic reinforcement learning with dual reward feedback, which combines execution-based and video-based rewards. Additionally, we design an annealing training strategy that gradually reduces the execution reward over time, allowing the model to rely more on the video reward in the later training stages. Experiments show that Video-Coder achieves significant gains, improving Minimax-H3 by approximately 7 points on PhyGenBench and 24 points on VBench-2.0. We hope this work can serve as an open foundation for coding agents in video generation, and we fully open-source our data, models, and code