https://www.dbpia.co.kr/author/authorDetail?ancId=671381487

Abstract

Keyframe-based motion graphics design is a time-consuming and labor-intensive process, often leading to reduced efficiency and increased fatigue for designers (Jahanlou et al., 2021). While scripts and expressions can automate repetitive tasks, they pose significant technical barriers for designers lacking coding experience (Sturman, 1998). Although recent AI-driven approaches like motion style transfer and data-based generation have emerged, they often overlook the interactive and creative process of expression authoring. This study proposes a novel system that integrates Large Language Models (ChatGPT API) and block-based programming to enable designers to easily create and control expressions. Specifically, the system (1) translates natural language intentions into expression codes via LLM and (2) provides a block-based interface for intuitive visualization and modification. Adopting a mixed-methods research design, we identified core pain points through literature review and in-depth interviews with motion designers, including tasks utilizing ChatGPT. Based on these findings, we designed an interaction system that merges LLM capabilities with block-based programming. This research contributes a new approach to Creativity Support Tools (CST) by empirically identifying design factors for non-programmers and exploring the potential of human-AI collaboration in motion graphics. The findings offer practical insights into a new paradigm of natural language-based interaction within creative design workflows.