EasyCache

KO | EN

gitlite — search

#javascript #ai #golang #android #machine-learning #deep-learning #linux #docker #nodejs #java #cli #security #ios #open-source

(★ 299)

Less is Enough: Training-Free Video Diffusion Acceleration via Runtime-Adaptive Caching

File Explorer

Download Latest Version (.zip)

- - 6.gif
  - 7.gif
- - 6.gif
  - 7.gif
- - 6.gif
  - 7.gif
- - 6.gif
  - 7.gif
- - video_metrics.py
- - baseline_544p.gif
  - baseline_720p.gif
  - easycache_544p.gif
  - svg_with_easycache_720p.gif
- easycache_sample_video.py
- hyvideo_svg_easycache.py
- README.md
- - grogu.png
- - video_metrics.py
- - i2v_easycache_14b_720p.gif
  - i2v_gt_14b_720p.gif
  - t2v_easycache_14b_720p.gif
  - t2v_gt_14b_720p.gif
- easycache_generate.py
- README.md
- - xwing.png
- - i2v-A14B-gt.gif
  - i2v-A14B-ours.gif
  - t2v-A14B-gt.gif
  - t2v-A14B-ours.gif
  - ti2v-5B-i2v-gt.gif
  - ti2v-5B-i2v-ours.gif
  - ti2v-5B-t2v-gt.gif
  - ti2v-5B-t2v-ours.gif
- easycache_generate.py
- README.md
.gitignore
LICENSE
README.md

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<div align="center"> 
 <h1>Less is Enough: Training-Free Video Diffusion Acceleration via Runtime-Adaptive Caching</h1>

<a href="https://lmd0311.github.io/" target="_blank" rel="noopener noreferrer">Xin Zhou</a>1\*,
 <a href="https://dk-liang.github.io/" target="_blank" rel="noopener noreferrer">Dingkang Liang</a>1\*,
Kaijin Chen1, Tianrui Feng1,
 <a href="https://scholar.google.com/citations?user=PVMQa-IAAAAJ&hl=en" target="_blank" rel="noopener noreferrer">Xiwu Chen</a>2, Hongkai Lin1, 
 <a href="https://scholar.google.com/citations?user=gdP9StQAAAAJ&hl=en" target="_blank" rel="noopener noreferrer">Yikang Ding</a>2, Feiyang Tan2,
 <a href="https://scholar.google.com/citations?user=4uE10I0AAAAJ&hl=en" target="_blank" rel="noopener noreferrer">Hengshuang Zhao</a>3,
 <a href="https://scholar.google.com/citations?user=UeltiQ4AAAAJ&hl=en" target="_blank" rel="noopener noreferrer">Xiang Bai</a>1†

1 Huazhong University of Science and Technology, 2 MEGVII Technology, 3 The University of Hong Kong

(\*) Equal contribution. (†) Corresponding author.

[![arXiv](https://img.shields.io/badge/Arxiv-2507.02860-b31b1b.svg?logo=arXiv)](https://arxiv.org/abs/2507.02860)
  [![Project](https://img.shields.io/badge/Homepage-project-orange.svg?logo=googlehome)](https://H-EmbodVis.github.io/EasyCache/)
  [![Code License](https://img.shields.io/badge/Code%20License-Apache_2.0-green.svg)](https://github.com/LMD0311/EasyCache/blob/main/LICENSE)

</div>

## 🎬 Visual Comparisons
Video synchronization issues may occur due to network load, for improved visualization, see the [project page](https://H-EmbodVis.github.io/EasyCache/)

**Prompt: "Grassland at dusk, wild horses galloping, golden light flickering across manes."**
*(HunyuanVideo)*

**Prompt: "A top-down view of a barista creating latte art, skillfully pouring milk to form the letters 'TPAMI' on coffee."**
*(Wan2.1-14B)*

**Compatibility with SVG**

SVG with EasyCache on HunyuanVideo can achieve more than 3x speedup.

https://github.com/user-attachments/assets/248ab88f-dfa8-4980-9b51-5c081e27db9a

## 📰 News
- **If you like our project, please give us a star ⭐ on GitHub for the latest update.**
- **[2025/08/01]** 🔥 EasyCache for [**Wan2.2**](https://github.com/H-EmbodVis/EasyCache/tree/main/EasyCache4Wan2.2) I2V is released.
- **[2025/07/31]** 🔥 EasyCache for [**Wan2.2**](https://github.com/H-EmbodVis/EasyCache/tree/main/EasyCache4Wan2.2) T2V-A14B is released.
- **[2025/07/30]** 🔥 EasyCache for [**Wan2.2**](https://github.com/H-EmbodVis/EasyCache/tree/main/EasyCache4Wan2.2) TI2V-5B is released. Thanks [@AChowdhury1211](https://github.com/AChowdhury1211).
- **[2025/07/06]** 🔥 EasyCache for [**Wan2.1**](https://github.com/H-EmbodVis/EasyCache/tree/main/EasyCache4Wan2.1) I2V is released.
- **[2025/07/05]** 🔥 EasyCache for [**Wan2.1**](https://github.com/H-EmbodVis/EasyCache/tree/main/EasyCache4Wan2.1) T2V is released.
- **[2025/07/04]** 🎉 Release the [**paper**](https://arxiv.org/abs/2507.02860) of EasyCache.
- **[2025/07/03]** 🔥 EasyCache for Sparse-VideoGen on [**HunyuanVideo**](https://github.com/H-EmbodVis/EasyCache/tree/main/EasyCache4HunyuanVideo) is released.
- **[2025/07/02]** 🔥 EasyCache for [**HunyuanVideo**](https://github.com/H-EmbodVis/EasyCache/tree/main/EasyCache4HunyuanVideo) is released.

## Abstract
Video generation models have demonstrated remarkable performance, yet their broader adoption remains constrained by slow inference speeds and substantial computational costs, primarily due to the iterative nature of the denoising process. Addressing this bottleneck is essential for democratizing advanced video synthesis technologies and enabling their integration into real-world applications. This work proposes EasyCache, a training-free acceleration framework for video diffusion models. EasyCache introduces a lightweight, runtime-adaptive caching mechanism that dynamically reuses previously computed transformation vectors, avoiding redundant computations during inference. Unlike prior approaches, EasyCache requires no offline profiling, pre-computation, or extensive parameter tuning. We conduct comprehensive studies on various large-scale video generation models, including OpenSora, Wan2.1, and HunyuanVideo. Our method achieves leading acceleration performance, reducing inference time by up to 2.1-3.3× compared to the original baselines while maintaining high visual fidelity with a significant up to 36% PSNR improvement compared to the previous SOTA method. This improvement makes our EasyCache a efficient and highly accessible solution for high-quality video generation in both research and practical applications.

## 🚀 Main Performance

We validated the performance of EasyCache on leading video generation models and compared it with other state-of-the-art training-free acceleration methods.

### Comparison with SOTA Methods

Tested on Vbench prompts with NVIDIA A800.

**Performance on HunyuanVideo:**
| Method | Latency (s)↓ | Speedup ↑ | PSNR ↑ | SSIM ↑ | LPIPS ↓ |
|:---:|:---:|:---:|:---:|:---:|:---:|
| HunyuanVideo (Baseline) | 1124.30 | 1.00x | - | - | - |
| PAB | 958.23 | 1.17x | 18.58 | 0.7023 | 0.3827 |
| TeaCache | 674.04 | 1.67x | 23.85 | 0.8185 | 0.1730 |
| SVG | 802.70 | 1.40x | 26.57 | 0.8596 | 0.1368 |
| **EasyCache (Ours)** | **507.97** | **2.21x** | **32.66** | **0.9313** | **0.0533** |

**Performance on Wan2.1-1.3B:**

| Method | Latency (s)↓ | Speedup ↑ | PSNR ↑ | SSIM ↑ | LPIPS ↓ |
|:---:|:---:|:---:|:---:|:---:|:---:|
| Wan2.1 (Baseline) | 175.35 | 1.00x | - | - | - |
| PAB | 102.03 | 1.72x | 18.84 | 0.6484 | 0.3010 |
| TeaCache | 87.77 | 2.00x | 22.57 | 0.8057 | 0.1277 |
| **EasyCache (Ours)** | **69.11** | **2.54x** | **25.24** | **0.8337** | **0.0952** |

### Compatibility with Other Acceleration Techniques

EasyCache is orthogonal to other acceleration techniques, such as the efficient attention mechanism SVG, and can be combined with them for even greater performance gains.

**Combined Performance on HunyuanVideo (720p):**
*Tested on NVIDIA H20 GPUs.*
| Method | Latency (s)↓ | Speedup ↑ | PSNR (dB) ↑ |
|:---:|:---:|:---:|:---:|
| Baseline | 6594s | 1.00x | - |
| SVG | 3474s | 1.90x | 27.56 |
| SVG (w/ TeaCache) | 2071s | 3.18x | 22.65 |
| SVG (w/ **Ours**) | **1981s** | **3.33x** | **27.26** |

## 🛠️ Usage
Detailed instructions for each supported model are provided in their respective directories. We are continuously working to extend support to more models.

## 🎯 To Do

- [x] Support HunyuanVideo
- [x] Support Sparse-VideoGen on HunyuanVideo
- [x] Support Wan2.1 T2V-1.3B/14B
- [x] Support Wan2.1 I2V-14B
- [x] Support Wan2.2 TI2V-5B
- [X] Support Wan2.2 T2V-A14B
- [X] Support Wan2.2 I2V

## 🌹 Acknowledgements
We would like to thank the contributors to the [Wan2.1](https://github.com/Wan-Video/Wan2.1), [HunyuanVideo](https://github.com/Tencent-Hunyuan/HunyuanVideo), [Wan2.2](https://github.com/Wan-Video/Wan2.2), [OpenSora](https://github.com/hpcaitech/Open-Sora), and [SVG](https://github.com/svg-project/Sparse-VideoGen) repositories, for their open research and exploration.
- [ComfyUI-WanVideoWrapper](https://github.com/kijai/ComfyUI-WanVideoWrapper) supports EasyCache4Wan2.1/EasyCache4Wan2.2. Thanks [@kijai](https://github.com/kijai).
- [ComfyUI](https://github.com/comfyanonymous/ComfyUI) supports EasyCache for several model includes Kontext, Qwen Image Edit, Wan2.2 14B. Thanks [@Kosinkadink](https://github.com/Kosinkadink).

## 📖 Citation

If you find this repository useful in your research, please consider giving a star ⭐ and a citation.
```bibtex
@article{zhou2025easycache,
  title={Less is Enough: Training-Free Video Diffusion Acceleration via Runtime-Adaptive Caching},
  author={Zhou, Xin and Liang, Dingkang and Chen, Kaijin and Feng, Tianrui and Chen, Xiwu and Lin, Hongkai and Ding, Yikang and Tan, Feiyang and Zhao, Hengshuang and Bai, Xiang},
  journal={arXiv preprint arXiv:2507.02860},
  year={2025}
}
```

EasyCache

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

EasyCache

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

Download README