SpeedVFI: One-step Diffusion for Efficient Video Frame Interpolation

Ganggui Ding, Xiaogang Xu, Hao Chen, Chunhua Shen
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:25263-25281, 2026.

Abstract

Generative video diffusion models have shown strong robustness to large motion and occlusions for video frame interpolation (VFI). However, their inference efficiency lags significantly behind learning-based methods due to the structural redundancy of pairwise inference and the procedural latency of multi-step iterative denoising. To address these limitations, we propose SpeedVFI, a task-specific one-step diffusion formulation that recasts generative VFI as unified sequence interpolation. SpeedVFI achieves dual efficiency improvements by interpolating the entire video sequence in a single forward pass to eliminate pairwise overhead, and by distilling the generation trajectory into a one-step denoising process to bypass iterative latency. To make this formulation effective for VFI, we introduce temporal RoPE alignment for temporally consistent conditioning and noise-centric partial attention to reduce computational overhead while preserving global context. Extensive experiments demonstrate that SpeedVFI accelerates diffusion-based VFI by orders of magnitude while maintaining competitive quantitative and visual quality.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-ding26t, title = {{S}peed{VFI}: One-step Diffusion for Efficient Video Frame Interpolation}, author = {Ding, Ganggui and Xu, Xiaogang and Chen, Hao and Shen, Chunhua}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {25263--25281}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/ding26t/ding26t.pdf}, url = {https://proceedings.mlr.press/v306/ding26t.html}, abstract = {Generative video diffusion models have shown strong robustness to large motion and occlusions for video frame interpolation (VFI). However, their inference efficiency lags significantly behind learning-based methods due to the structural redundancy of pairwise inference and the procedural latency of multi-step iterative denoising. To address these limitations, we propose SpeedVFI, a task-specific one-step diffusion formulation that recasts generative VFI as unified sequence interpolation. SpeedVFI achieves dual efficiency improvements by interpolating the entire video sequence in a single forward pass to eliminate pairwise overhead, and by distilling the generation trajectory into a one-step denoising process to bypass iterative latency. To make this formulation effective for VFI, we introduce temporal RoPE alignment for temporally consistent conditioning and noise-centric partial attention to reduce computational overhead while preserving global context. Extensive experiments demonstrate that SpeedVFI accelerates diffusion-based VFI by orders of magnitude while maintaining competitive quantitative and visual quality.} }
Endnote
%0 Conference Paper %T SpeedVFI: One-step Diffusion for Efficient Video Frame Interpolation %A Ganggui Ding %A Xiaogang Xu %A Hao Chen %A Chunhua Shen %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-ding26t %I PMLR %P 25263--25281 %U https://proceedings.mlr.press/v306/ding26t.html %V 306 %X Generative video diffusion models have shown strong robustness to large motion and occlusions for video frame interpolation (VFI). However, their inference efficiency lags significantly behind learning-based methods due to the structural redundancy of pairwise inference and the procedural latency of multi-step iterative denoising. To address these limitations, we propose SpeedVFI, a task-specific one-step diffusion formulation that recasts generative VFI as unified sequence interpolation. SpeedVFI achieves dual efficiency improvements by interpolating the entire video sequence in a single forward pass to eliminate pairwise overhead, and by distilling the generation trajectory into a one-step denoising process to bypass iterative latency. To make this formulation effective for VFI, we introduce temporal RoPE alignment for temporally consistent conditioning and noise-centric partial attention to reduce computational overhead while preserving global context. Extensive experiments demonstrate that SpeedVFI accelerates diffusion-based VFI by orders of magnitude while maintaining competitive quantitative and visual quality.
APA
Ding, G., Xu, X., Chen, H. & Shen, C.. (2026). SpeedVFI: One-step Diffusion for Efficient Video Frame Interpolation. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:25263-25281 Available from https://proceedings.mlr.press/v306/ding26t.html.

Related Material