Large language models (LLMs) have the potential to impact a wide range of creative domains, as seen in popular text-to-image generators like DALL·E and Midjourney. However, the use of LLMs in motion-based visual design has not yet been explored and presents unique challenges, such as how users can effectively describe motion using natural language. Many current generative design tools also do not support the iterative refinement of designs beyond initial prompts.
In this study, we introduce Keyframer, a design tool that utilizes the code generation capabilities of LLMs to facilitate design exploration for animations. Keyframer was developed based on insights gathered from interviews with professional motion designers, animators, and engineers. The tool is designed to assist users in both the ideation and refinement stages of animation design processes by allowing them to experiment with different design variations throughout their workflow.
We conducted a user evaluation with 13 participants of varying animation and programming backgrounds to understand their prompting strategies and how they integrated design variants into their workflow using Keyframer. Through this evaluation, we identified a set of design principles for incorporating LLMs into motion design prototyping tools and discussed their potential impact on visual design tools as a whole.
