[ECCV 2024] StoryImager: A Unified and Efficient Framework for Coherent Story Visualization and Completion

A high-quality, unified, and efficient framework for story visualization and completion

Official Pytorch implementation for our paper StoryImager: A Unified and Efficient Framework for Coherent Story Visualization and Completion by Ming Tao, Bing-Kun Bao, Hao Tang, Yaowei Wang, Changsheng Xu.

Requirements

• python 3.9
• Pytorch 1.13

Preparation

Datasets

1. Download the preprocessed data for PororoSV FlintstonesSV and extract them to data/

Training

Evaluation

Download Pretrained Model

Sampling

Synthesize images from your story descriptions

• the sample.ipynb can be used to sample

---

Citing StoryImager

If you find StoryImager useful in your research, please consider citing our paper:

@article{tao2024storyimager,
  title={StoryImager: A Unified and Efficient Framework for Coherent Story Visualization and Completion},
  author={Tao, Ming and Bao, Bing-Kun and Tang, Hao and Wang, Yaowei and Xu, Changsheng},
  journal={arXiv preprint arXiv:2404.05979},
  year={2024}
}