A visual anlaytics system for exploring prompt history when using text-to-image generative models (online demo, paper).
Client and server construct the web interface. Diffusion and preprocess generate, log, and preprocess images.
Currently, the backends (diffusion, preprocess, and server) are hosted at https://vis.pku.edu.cn/prompthis. The host is sometimes inaccessible due to domain restrictions. Details on setting up the backends locally are coming soon.
To use the front end,
cd client
- Run
npm install
to install dependencies - Run
npm run dev
to launch the dev server
@article{guo2024prompthis,
author={Guo, Yuhan and Shao, Hanning and Liu, Can and Xu, Kai and Yuan, Xiaoru},
journal={IEEE Transactions on Visualization and Computer Graphics},
title={PrompTHis: Visualizing the Process and Influence of Prompt Editing during Text-to-Image Creation},
year={2024},
doi={10.1109/TVCG.2024.3408255}
}