4Director

4Director: Controlling Video World Models with Rigid 3D Geometry

4Director conditions a video world model on an explicit 4D scene: one input image becomes an editable scene of a background point cloud and object meshes, and every control (moving or turning an object, inserting a new one, orbiting the camera) is a rigid 3D transform applied inside that scene. The scene is rendered into a depth control video, which this model turns into the final video.

Please note: For individuals or organizations generating annual revenue of US $1,000,000 (or local currency equivalent) or more, regardless of the source of that revenue, you must obtain an enterprise commercial license directly from Stability AI before commercially using 4Director, derivative works of 4Director, or outputs from 4Director. See https://stability.ai/license and https://stability.ai/enterprise.

Model Description

• Developed by: Stability AI and the University of Illinois at Urbana-Champaign
• Authors: Wei Cao, Hao Zhang, Vikram Voleti, Yuqun Wu, Mallikarjun B R, Shimon Vainer, Mark Boss, Yaoyao Liu
• Model type: the Motion Adapter, the fine-tuned VACE weights for Wan2.1-VACE-14B
• Inputs: a reference image, a depth control video rendered from the 4D scene, and a text prompt
• Output: an 81-frame, 832×480 video at 16 fps (20 sampling steps by default)

Model Sources

• Repository: https://github.com/VVeiCao/4Director
• Paper: https://arxiv.org/abs/2610.02160
• Project page: https://stability-ai.github.io/4director/

Usage

Follow the GitHub repository to install the environment, then:

git clone --recurse-submodules https://github.com/VVeiCao/4Director.git && cd 4Director
python download_models.py   # SAM 2, Wan2.1-VACE-14B, and this checkpoint, into ./models
python run.py --config configs/cases/object-motion.yaml

Files

• step-2610.safetensors: the Motion Adapter, SHA-256 91c56861c58dfa618b358d80093bee8561c082844fd215c968b00eadb132c9e3. download_models.py checks it.

License

• Community License: Free for research, non-commercial, and commercial use by organizations and individuals generating annual revenue of US $1,000,000 (or local currency equivalent) or less, regardless of the source of that revenue.
• If your annual revenue exceeds US $1M, any commercial use of this model or derivative works requires an Enterprise License directly from Stability AI.

The full text is in LICENSE.md; see also https://stability.ai/license. The base model, Wan2.1-VACE-14B, is released under its own license.

Intended Uses

Intended uses include research on controllable video generation and video world models, in particular on camera and multi-object motion control through explicit 3D geometry, and reproduction of the paper's results.

Out-of-Scope Uses

4Director is not intended to generate factual representations of people, events, products, or places. Use must comply with Stability AI's Acceptable Use Policy and license terms.

Citation

@misc{cao20264directorcontrollingvideoworld,
  title={4Director: Controlling Video World Models with Rigid 3D Geometry},
  author={Wei Cao and Hao Zhang and Vikram Voleti and Yuqun Wu and Mallikarjun B R and Shimon Vainer and Mark Boss and Yaoyao Liu},
  year={2026},
  eprint={2610.02160},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2610.02160},
}

Safety

Users should evaluate generated outputs for their own application constraints and apply additional mitigations where needed. Report safety issues to safety@stability.ai and security issues to security@stability.ai.