README.md

metadata

base_model:
  - Wan-AI/Wan2.1-T2V-1.3B
  - Qwen/Qwen-Image-Edit-2509
language:
  - en
license: apache-2.0
pipeline_tag: other
library_name: diffusers
arxiv: 2601.20175
tags:
  - video
  - image
  - stylization
  - style-transfer

TeleStyle: Content-Preserving Style Transfer in Images and Videos

Shiwen Zhang, Xiaoyan Yang, Bojia Zi, Haibin Huang, Chi Zhang, Xuelong Li

Institute of Artificial Intelligence, China Telecom (TeleAI)  

     

TeleStyle is a lightweight yet effective model for both image and video stylization. Built upon Qwen-Image-Edit, it leverages a Curriculum Continual Learning framework to achieve high-fidelity content preservation and style customization across diverse, in-the-wild style categories.

🔔 News

• [2026-01-28]: Released Code, Model, Paper.

How to use

1. Installation

pip install -r requirements.txt

2. Inference

We provide inference scripts for running TeleStyle on demo inputs for each task:

Image Stylization

To generate a stylized image using a reference style image and a content image:

python telestyleimage_inference.py --image_path assets/example/0.png --style_path videos/1.png --output_path results/image.png

Video Stylization

To generate a stylized video using a stylized first frame and a content video:

python telestylevideo_inference.py --video_path assets/example/1.mp4 --style_path assets/example/1-0.png --output_path results/video.mp4

For more details, please refer to the 🔗 GitHub repository.

🌟 Citation

If you find TeleStyle useful for your research and applications, please cite using this BibTeX:

@article{teleai2026telestyle,
    title={TeleStyle: Content-Preserving Style Transfer in Images and Videos}, 
    author={Shiwen Zhang and Xiaoyan Yang and Bojia Zi and Haibin Huang and Chi Zhang and Xuelong Li},
    journal={arXiv preprint arXiv:2601.20175},
    year={2026}
}