Combining direct manipulation and prompting in AI-powered animation tools

Determine how to design AI-powered animation tools that combine direct manipulation interfaces (e.g., timeline-based or systems like DirectGPT) with natural language prompting to streamline the definition and editing of animated properties.

Background

The paper observes that many animators are accustomed to timeline-based, direct manipulation tools (e.g., After Effects) and that different tasks may benefit from different interaction modes: transforms can be faster via direct manipulation, while more complex or non-translational animations may be easier through prompting.

Building on Keyframer’s prompt-driven workflow with code/property editors, the authors identify a specific open question about how best to blend direct manipulation with prompting to achieve efficient and precise control over animated properties.