A Generative Motion Rig for Artist-Driven Motion Authoring

How to read this

Category

SIGGRAPH Talk describing a production tool, a generative motion authoring rig built as an animation software plugin

Contributions

• Wraps a general motion generation model in a familiar rig interface, a posing rig plus sparse handles onto Neural Motion Curves, so animators keep a click and drag workflow
• Introduces a generative keyframing workflow where artists set sparse poses, handles, window length and noise sampling to generate and resample whole motion sequences
• Extends the same rig to generative motion editing, letting artists edit and extend existing mocap clips rather than only generating from scratch
• Uses a client server split, the generative model runs on a dedicated GPU server while the DCC (Maya, Blender) stays a thin client, so the same backend can plug into different host software

Context

This is a direct production wrapper around SKEL-Betweener: a Neural Motion Rig for Interactive Motion Authoring (2024), reusing its Neural Motion Curves as the sparse handle representation and adding an artist facing rig plus plugin on top. Where the 2024 work introduced the interactive motion model itself, this talk is about turning that model into a tool an animator can sit down with, including workflow lessons on bridging generative and traditional animation.Builds on: SKEL-Betweener: a Neural Motion Rig for Interactive Motion Authoring

Correctness

The claims rest on a demonstration video and artist feedback from internal use, not a quantitative study, this is a 3 page SIGGRAPH Talk rather than a full paper, so there is no formal user study or benchmark. Treat the speedup and quality claims as qualitative production experience rather than measured results.

Clarity

Short and readable, written for riggers and pipeline TDs rather than ML researchers, the underlying motion model's math is left to the 2024 paper it cites.

How to read it

First pass, read the abstract and the client server diagram to understand the split between DCC plugin and GPU hosted model. Second pass, read the generative keyframing and generative motion editing sections to see which controls, handles, window length, noise, map to which artist actions. Third pass is only worth it if you are building a similar in house generative rig, in which case also read the cited SKEL-Betweener paper for the underlying motion representation.