Notational Animating: An Interactive Approach to Creating and Editing Animation Keyframes

Paper Title

Notational Animating: An Interactive Approach to Creating and Editing Animation Keyframes

Publication Info

• Topic area: Animation keyframe authoring using sketched notations and generative AI.
• Keywords: Animation keyframes, sketched notations, generative AI, vision-language models, animation tools, user-defined abstractions, motion dynamics, interactive design, animation principles, feedback mechanisms.

Background and Problem

• Problem / challenge: Existing animation tools rely on rigid, system-defined abstractions that limit flexibility and generalizability. They often fail to accommodate the contextual, ambiguous, and combinational nature of animators' traditional notations.
• Significance: Addressing this gap can make animation authoring more intuitive, reduce the cognitive load on animators, and streamline the creation of expressive, nuanced animations.
• Motivation and related work: Prior tools like Draco, Kitty, and Motion Doodle map predefined visual abstractions to specific animation effects, but they enforce fixed vocabularies and lack support for user-defined, improvisational notations. Generative AI tools offer high-quality outputs but provide limited fine-grained control. This paper builds on these limitations to propose a more flexible, user-centered approach.

Solution

• Proposed approach: Notational animating—an interaction paradigm where animators sketch high-level notations over static drawings to indicate motion, which are interpreted by vision-language models (VLMs) to generate animation keyframes.
• Novelty:
  1. Characterization of animators’ notational practices and their contextual, ambiguous, and combinational nature.
  2. Formalized animation representation using ⟨source, path, target⟩ tuples to interpret informal notations systematically.
  3. A prototype system with a two-level feedback mechanism and dynamic UI widgets for fine-grained control.
  4. Insights from a preliminary expert study on how professionals perceive and use notational animating.
• Procedure and key techniques:
  1. Analyze 135 real-world animator sketches to identify recurring notation patterns.
  2. Develop a structured animation representation (⟨source, path, target⟩) for systematic interpretation of notations.
  3. Implement a prototype system with:
     • A drawing canvas for sketching and notating.
     • Vision-language models to interpret notations and generate keyframes.
     • Feedback mechanisms (motion tags, timelines, sliders) for iterative refinement.
  4. Conduct a user study with 7 professional animators to evaluate usability and effectiveness.

Results

• Concrete findings:
  • Participants found the system intuitive, with 6/7 agreeing it was easy to express animation intent.
  • Dynamic sliders were used approximately 8.1 times per session for fine-tuning motion range and intensity.
  • Observed notating patterns included global vibes, target poses, hierarchical dynamism, and individual pieces.
• Advantage over baselines:
  • Supports user-defined, flexible notations rather than rigid system-defined abstractions.
  • Enables holistic keyframe authoring by combining high-level sketching with low-level parameter tuning.
  • Facilitates iterative refinement with structured feedback and dynamic UI widgets.
• Experiments / evaluation:
  • Conducted with 7 professional animators (2 females, 5 males, aged 26–46, with 2–20+ years of experience).
  • Tasks included constrained (targeted effects) and exploratory (open-ended) animation creation.
  • Feedback was analyzed through thematic analysis of think-aloud data and interviews.
• Limitations and future work:
  • Current system struggles with 3D spatial changes, fine-grained occlusion, and non-geometric transformations (e.g., lighting).
  • AI models occasionally misinterpret intent or fail to execute exaggerated deformations.
  • Future work includes faster inference, improved generative fidelity, and extending notations to camera movements and non-geometric effects.

Summary

This paper introduces notational animating, a novel paradigm for animation keyframe authoring that leverages user-defined sketched notations interpreted by vision-language models. The approach addresses limitations in existing tools by supporting contextual, ambiguous, and combinational notations, enabling animators to express nuanced motion dynamics intuitively. A prototype system demonstrates the feasibility of this concept, with structured feedback and dynamic UI widgets facilitating iterative refinement. Preliminary expert evaluations highlight its potential to complement traditional animation workflows, though challenges remain in handling 3D spatial changes and improving generative fidelity. The work opens new directions for AI-assisted animation tools that align with animators’ creative processes.