Key points
- Animaj researchers presented a learned in-betweening method at SIGGRAPH 2026 that generates dense 3D animation from sparse blocking poses inside Autodesk Maya.
- Its Adaptive Interpolation-Synthesis layer dynamically balances learned interpolation against direct pose synthesis, accelerating the in-betweening stage by up to 3.5 times.
- The paper is the studio's first-party technical account of the AI in-betweening stage in the pipeline behind its productions, including Disney Jr's Ozzy Fox.
Summary
A SIGGRAPH 2026 conference paper from Animaj (Paris) describing the machine learning method behind the studio's production in-betweening tool. The Adaptive Interpolation-Synthesis (AIS) layer, combined with a bidirectional LSTM encoder, mirrors the animator's process by dynamically blending learned interpolation with direct pose synthesis, and a domain-based input keypose schedule aligns training with real production data. Integrated into Autodesk Maya, the studio reports in-betweening completed up to 3.5 times faster while preserving motion style. In-betweening is a foundational taught skill, and this is a rare production-proven, peer-reviewed account of restructuring that stage around machine learning while keeping the animator's blocking-to-polish workflow.
Related items
- Disney Jr launches Ozzy Fox, its first series made with AI-native studio Animaj
- A deep learning framework for character motion synthesis and editing (SIGGRAPH 2016 / SIGGRAPH 2026 Test-of-Time award)
Source
Source: SIGGRAPH 2026 Conference Papers / arXiv ↗ (Research)
Cite this item
Anton Raël, Julien Boucher & Antoine Lhermitte (2026). ‘Adaptive Interpolation-Synthesis for Motion In-Betweening on Keyframe-Based Animation’, SIGGRAPH 2026 Conference Papers / arXiv. Available at: https://arxiv.org/abs/2605.02742
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.