← ArchiveTalk2024
Deliver Faster Rigging and Animation with AI
Autodesk Animation Product Manager Lance Thornton and Sr. Principal Research Scientist Evan Atherton explore how AI and ML techniques can accelerate rigging and animation production workflows in Maya.
Abstract
Autodesk's Lance Thornton and Evan Atherton present two AI approaches to character work in Maya. Thornton demonstrates the machine learning deformer, which trains locally on automatically generated or animation-driven poses to approximate a complex deformer or chain (such as a hero muscle system) with a single constant-time node, showing roughly 4x faster playback and interactive manipulation on a Digital Fish rig, support for bodies and faces, paintable weights and falloffs, CPU or GPU evaluation, switching back to the full rig for polish, and a mesh-compare heat map. Atherton then shows a neural motion control research prototype built in Bifrost, where a small neural network trained on under an hour of motion capture predicts each next pose from the current pose and a keyframed trajectory, producing art-directable walk, run, jump, and sit behaviors for quadrupeds and bipeds. The motion is passed through Human IK to retarget behavior across rigs of differing proportions and bake to standard controllers, with animation layers for final tweaks and Bifrost particle instancing for crowds.
How to read this
- Category
- Production talk: AI/ML for rigging and animation in Maya
- Contributions
- Demonstrates a machine learning deformer that trains locally to approximate a complex deformer or chain with a single constant-time node, with faster playback, paintable weights/falloffs, CPU or GPU evaluation, and a mesh-compare heat map
- Shows a neural motion control prototype in Bifrost where a small network trained on under an hour of mocap predicts the next pose from the current pose and a keyframed trajectory
- Covers art-directable walk/run/jump/sit for quadrupeds and bipeds, with Human IK retargeting across rigs and animation layers for final polish
- Context
- Brings learned-deformer and learned-motion ideas into a commercial DCC, in the spirit of fast deformation approximation work such as Bailey et al.'s 'Fast and Deep Deformation Approximations'.Builds on: Fast and Deep Deformation Approximations
- Correctness
- Studio/vendor practice, not peer-reviewed; results are demo-proven on specific rigs (e.g. a Digital Fish rig) and described as approximations of the full rig, with the full rig retained for polish, so quality and the cited speedups are illustrative rather than benchmarked.
- Clarity
- Very accessible; a single viewing conveys both tools and their intended workflow role.
- How to read it
- Watch once to understand where learned deformers and neural motion fit in a Maya pipeline and their tradeoffs (approximation vs full-rig polish); no deep formal pass needed, treat as a capabilities demo.
Builds on
Built upon by
Nothing yet.
Related work
- Mobilizing Mocap, Motion Blending, and Mayhem: Rig Interoperability for Crowd Simulation on Incredibles 2 2018 / SIGGRAPH
- One Model to Rig Them All: Diverse Skeleton Rigging with UniRig 2025 / SIGGRAPH
- Motion Retargeting for Crowd Simulation 2015 / DigiPro
- MoRig: Motion-Aware Rigging of Character Meshes from Point Clouds 2022 / SIGGRAPH Asia
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →