CSGO: Content-Style Composition in Text-to-Image Generation

Official 🤗 Gradio demo for CSGO: Content-Style Composition in Text-to-Image Generation.
How to use:

  1. Upload a content image if you want to use image-driven style transfer.
  2. Upload a style image.
  3. Sets the type of task to perform, by default image-driven style transfer is performed. Options are Image-driven style transfer, Text-driven style synthesis, and Text editing-driven style synthesis.
  4. If you choose a text-driven task, enter your desired prompt.
  5. If you don't provide a prompt, the default is to use the BLIP model to generate the caption. We suggest that by providing detailed prompts for Content images, CSGO is able to effectively guarantee content.
  6. Click the Submit button to begin customization.
  7. Share your stylized photo with your friends and enjoy! 😊

Advanced usage:

  1. Click advanced options.
  2. Choose different guidance and steps.
task
prompt_type

Choose to enter more detailed prompts yourself or use the blip model to describe content images.
0 2
0 2
1 15
1 4
5 100
-1000000 2147483647
Examples
Content Image (optional) Style Image task Prompt Content Scale Style Scale guidance scale Seed Value

📝 Tips In CSGO, the more accurate the text prompts for content images, the better the content retention. Text-driven style synthesis and text-edit-driven style synthesis are expected to be more stable in the next release.

📝 Citation
If our work is helpful for your research or applications, please cite us via:

@article{xing2024csgo,
       title={CSGO: Content-Style Composition in Text-to-Image Generation}, 
       author={Peng Xing and Haofan Wang and Yanpeng Sun and Qixun Wang and Xu Bai and Hao Ai and Renyuan Huang and Zechao Li},
       year={2024},
       journal = {arXiv 2408.16766},
}

📧 Contact
If you have any questions, please feel free to open an issue or directly reach us out at xingp_ng@njust.edu.cn.