For text-to-image generation, automatically refining user-provided natural language prompts into the keyword-enriched prompts favored by systems is essential for the user experience. Such a prompt refinement process is analogous to translating the prompt from "user languages" into "system languages". However, the scarcity of such parallel corpora makes it difficult to train a prompt refinement model. Inspired by zero-shot machine translation techniques, we introduce Prompt Refinement with Image Pivot (PRIP). PRIP innovatively uses the latent representation of a user-preferred image as an intermediary "pivot" between the user and system languages. It decomposes the refinement process into two data-rich tasks: inferring representations of user-preferred images from user languages and subsequently translating image representations into system languages. Thus, it can leverage abundant data for training. Extensive experiments show that PRIP substantially outperforms a wide range of baselines and effectively transfers to unseen systems in a zero-shot manner.
@article{arxiv.2407.00247,
title = {Prompt Refinement with Image Pivot for Text-to-Image Generation},
author = {Jingtao Zhan and Qingyao Ai and Yiqun Liu and Yingwei Pan and Ting Yao and Jiaxin Mao and Shaoping Ma and Tao Mei},
journal= {arXiv preprint arXiv:2407.00247},
year = {2024}
}