Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.05811 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910175413141504 |
|---|---|
| author | Menn, Dennis Yang, Yuedong Wang, Bokun Wei, Xiwen Munir, Mustafa Liang, Feng Marculescu, Radu Xu, Chenfeng Marculescu, Diana |
| author_facet | Menn, Dennis Yang, Yuedong Wang, Bokun Wei, Xiwen Munir, Mustafa Liang, Feng Marculescu, Radu Xu, Chenfeng Marculescu, Diana |
| contents | Current video generation models suffer from high computational latency, making real-time applications prohibitively costly. In this paper, we address this limitation by exploiting the temporal redundancy inherent in video latent patches. To this end, we propose the Latent Inter-frame Pruning with Attention Recovery (LIPAR) framework, which detects and skips recomputing duplicated latent patches. Additionally, we introduce a novel Attention Recovery mechanism that approximates the attention values of pruned tokens, thereby removing visual artifacts arising from naively applying the pruning method. Empirically, our method increases video editing throughput by $1.53\times$, achieving an average of 19.3 FPS on an NVIDIA RTX 4090 with the 1.3B Self-Forcing model (4-step denoising, FP16). The proposed method does not compromise generation quality and can be seamlessly integrated with the model without additional training. Our approach effectively bridges the gap between traditional compression algorithms and modern generative pipelines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_05811 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Video Compression Meets Video Generation: Latent Inter-Frame Pruning with Attention Recovery Menn, Dennis Yang, Yuedong Wang, Bokun Wei, Xiwen Munir, Mustafa Liang, Feng Marculescu, Radu Xu, Chenfeng Marculescu, Diana Computer Vision and Pattern Recognition Current video generation models suffer from high computational latency, making real-time applications prohibitively costly. In this paper, we address this limitation by exploiting the temporal redundancy inherent in video latent patches. To this end, we propose the Latent Inter-frame Pruning with Attention Recovery (LIPAR) framework, which detects and skips recomputing duplicated latent patches. Additionally, we introduce a novel Attention Recovery mechanism that approximates the attention values of pruned tokens, thereby removing visual artifacts arising from naively applying the pruning method. Empirically, our method increases video editing throughput by $1.53\times$, achieving an average of 19.3 FPS on an NVIDIA RTX 4090 with the 1.3B Self-Forcing model (4-step denoising, FP16). The proposed method does not compromise generation quality and can be seamlessly integrated with the model without additional training. Our approach effectively bridges the gap between traditional compression algorithms and modern generative pipelines. |
| title | Video Compression Meets Video Generation: Latent Inter-Frame Pruning with Attention Recovery |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.05811 |