<div align="center"> <p><img src="assets/logo2.png" width="360px"></p> <h1>SparkVSR: Interactive Video Super-Resolution via Sparse Keyframe Propagation</h1> <p> Jiongze Yu¹, Xiangbo Gao¹, Pooja Verlani², Akshay Gadde², Yilin Wang², Balu Adsumilli², Zhengzhong Tu^†,1 </p> <p> ¹Texas A&amp;M University &nbsp;&nbsp; ²YouTube, Google <br> ^†Corresponding author </p> <p> <a href="https://sparkvsr.github.io/"><img src="https://img.shields.io/badge/Project-Page-Green"></a> &nbsp; <a href="https://huggingface.co/JiongzeYu/SparkVSR"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Model-blue"></a> &nbsp; <a href="https://arxiv.org/abs/2603.16864"><img src="https://img.shields.io/badge/arXiv-2603.16864-b31b1b.svg"></a> </p> </div>

💡 Your ⭐ star means a lot to us and helps support the continuous development of this project!

ComfyUI_SparkVSR_SM

---

SparkVSR: Interactive Video Super-Resolution via Sparse Keyframe Propagation

Update

• 修复一些错误，超大视频请按工作流的note来调整tile数值

• Preprocess the image node to connect to the pisa_sr node to load the model, or use other methods to upscale a single-frame image, that is, enable reference image mode; otherwise, it is non-reference image mode. ref_indices are the sequence frame numbers of the reference image in the source video. For example,' 0, 'is the first frame. You can also input 0 to only reference the first frame to save resources and achieve better results.

• 预处理图片节点接入pisa_sr 节点加载的模型 或者你用其他方法放大的单帧图片，即开启图模式，否则是非垫图模式，ref_indices是垫图在源视频的序列帧序号，比如0就是首帧，你也可以输入'0,'(注意有逗号) 只参考首帧以节省资源获得更好的效果.

1.Installation

In the ./ComfyUI/custom_nodes directory, run the following:

git clone https://github.com/smthemex/ComfyUI_SparkVSR_SM

2.requirements

pip install -r requirements.txt

3.checkpoints

• cog unet merged or gguf links #f32 or bf16 or gguf
• cog vae links
• if use pisa_sr need stable-diffusion-2-1-base vae and unet,and pisa_sr.pklgoogle

├── ComfyUI/models/
|     ├── diffusion_models/
|        ├──sd21base-f32.safetensors  # 3.22G or fb16  if use  pisa_sr to refer image  optional 可选，使用 pisa_sr超分参考图用
|        ├──SparkVSR-S2-F32.safetensors # 20.7G or fb16
|     ├── gguf/
|        ├──SparkVSR-S2-Q8_0.gguf optional 可选，
|     ├── vae/
|        ├──CogVideoX1.5-5B-I2V-VAE.safetensors #822M
|        ├──sd21vae.safetensors # 319M  if use  pisa_sr to refer image  optional 可选，使用 pisa_sr超分参考图用
|     ├── loras
|        ├──pisa_sr.pkl # 32M  if use  pisa_sr to refer image optional 可选,使用 pisa_sr超分参考图用

4.Example

5.Citation

@misc{yu2026sparkvsrinteractivevideosuperresolution,
      title={SparkVSR: Interactive Video Super-Resolution via Sparse Keyframe Propagation}, 
      author={Jiongze Yu and Xiangbo Gao and Pooja Verlani and Akshay Gadde and Yilin Wang and Balu Adsumilli and Zhengzhong Tu},
      year={2026},
      eprint={2603.16864},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2603.16864}, 
}