Key points
- Adapts live action video diffusion priors to drawn animation via toon rectification learning.
- Generates in between frames between two cartoon keyframes, including new content in the gap.
- Peer reviewed at SIGGRAPH Asia with open source code and weights.
Summary
ToonCrafter, published at SIGGRAPH Asia 2024, addresses the in-betweening problem in drawn animation by adapting live-action video diffusion priors to the drawn image domain through a technique called toon rectification learning. Given two cartoon keyframes, the model generates plausible intermediate frames including content not visible in either keyframe, filling the gap generatively. The paper is open-sourced with public weights, making it directly deployable in teaching labs. It matches the scope contract's exemplar class of SIGGRAPH papers on neural in-betweening.
Related items
- AnimateDiff: Animate Your Personalized Text-to-Image Models without Specific Tuning
- Iterative Motion Editing with Natural Language
- Align Your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
Source
Source: SIGGRAPH Asia 2024 ↗ (Research)
Cite this item
SIGGRAPH Asia 2024 (2024). ‘ToonCrafter: Generative Cartoon Interpolation’, SIGGRAPH Asia 2024. Available at: https://dl.acm.org/doi/abs/10.1145/3687761
Your reference manager can also read this page directly: with the Zotero (or Mendeley) browser connector installed, save it straight to your library. Whole-collection exports: RIS, BibTeX, CSL-JSON.