ScholarWorks@숙명여자대학교

초록

As the webtoon market has grown rapidly, there have been attempts to convert various text content into webtoons. However, research on automated conversion systems that understand the context of text and maintain visual consistency while considering webtoon characteristics remains limited. This paper proposes a prompt optimization system for automatic conversion of diverse text genres into webtoons. The proposed system is based on GPT-4 and DALL-E 3, implementing image evaluation and prompt enhancement mechanisms utilizing the CLIP model. This study differentiates itself from existing research by introducing a CLIP model-based quantitative evaluation method and implementing mechanisms to maintain visual consistency between sequential images.

키워드