Repository files navigation

Geometry Forcing: Marrying Video Diffusion and 3D Representation for Consistent World Modeling

Haoyu Wu$^{1*}$,Diankun Wu$^{2*}$, Tianyu He$^{1†}$, Junliang Guo$^{1}$, Yang Ye$^{1}$, Yueqi Duan$^{2}$, Jiang Bian$^{1}$

$^1$ Microsoft Research$^2$ Tsinghua University

($^*$ Equal Contribution. † Project Lead)

Geometry Forcing (GF) Overview.(a) Our proposed GF paradigm enhances video diffusion models by aligning with geometric features from VGGT.(b) Compared to DFoT, our method generates more temporally and geometrically consistent videos.(c) While baseline features fail to reconstruct meaningful 3D geometry, GF-learned features enable accurate 3D reconstruction.

• [2025/10/8] We release the evaluation code for reprojection error and revisit error.
• [2025/9/24] We release code and checkpoint.
• [2025/9/22]Geometry Forcing is accepted toNeurIPS 2025 NextVid Workshop as an Oral!
• [2025/7/10] We release the paper and the project.

conda create -n geometryforcing python=3.10 -yconda activate geometryforcingpip install -r requirements.txt

We use Weights & Biases for logging.Sign up if you don't have an account, andmodifywandb.entity inconfig.yaml to your user/organization name.

1. Download pretrained checkpiont using huggingface:

bash scripts/hf_download_checkpoints.sh

2. Download pretrained checkpiont using modelscope:

bash scripts/ms_download_checkpoints.sh

3. Download and process RealEstate10k dataset todata/real-estate-10k

1. Single Image to Long Video (256 Frames):

bash scripts/eval_geometry_forcing.sh

2. Single Image to Rotation Video (16 Frames):

bash scripts/eval_geometry_forcing_rotation.sh

To train Geometry Forcing, run the following command:

bash scripts/train_geometry_forcing.sh

To evaluate the reprojection error and revisit error, please follow the instructions inREADME_EVAL.md.

If you find our work useful for your research, please consider citing our paper:

@article{wu2025geometryforcing,  title={Geometry Forcing: Marrying Video Diffusion and 3D Representation for Consistent World Modeling},  author={Wu, Haoyu and Wu, Diankun and He, Tianyu and Guo, Junliang and Ye, Yang and Duan, Yueqi and Bian, Jiang},  journal={arXiv preprint arXiv:2507.07982},  year={2025}}