You can deploy this model and start earning money today!

JiaqiFeng/
OneForcing

text-to-video

2605.23458

Views

No views yet

One-Forcing: Towards Stable One-Step Autoregressive Video Generation

One-Forcing enables stable 1-step autoregressive video generation by augmenting DMD-based causal distillation with a shared noised-latent adversarial critic. It achieves state-of-the-art 1-step VBench performance and efficient framewise generation.

Project Page | Code | Paper

Introduction

Recent advances have substantially improved real-time interactive video generation in the autoregressive regime. One-Forcing addresses the quality degradation and latency issues in one-step settings by augmenting the DMD objective with an auxiliary GAN loss. Experiments show it establishes state-of-the-art performance among one-step causal video generation methods.

Inference

To use the model, follow the installation instructions in the official repository. You can run inference using the following commands:

bash

# Download the trained One-Forcing checkpoint
hf download JiaqiFeng/OneForcing checkpoints/one_forcing.pt --local-dir .

# Run the inference script
bash scripts/infer.sh \
  --checkpoint_path checkpoints/one_forcing.pt \
  --prompt_path prompts/demos.txt \
  --output_folder outputs

Citation

bibtex

@article{feng2026oneforcing,
  title={One-Forcing: Towards Stable One-Step Autoregressive Video Generation},
  author={Feng, Jiaqi and Cui, Justin and Ban, Yuanhao and Hsieh, Cho-Jui},
  journal={arXiv preprint arXiv:2605.23458},
  year={2026},
  eprint={2605.23458},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2605.23458}
}

Acknowledgements

This codebase builds on Causal Forcing, Self Forcing, CausVid, and the Wan model family.