Skip to content

← ArchivePaper2020

Robust Motion In-Betweening

Felix Harvey, Mike Yurick, Christopher Pal, Derek Nowrouzezahrai

SIGGRAPHUbisoft414 cites8 descendantsMotion Synthesis

Recurrent neural network for robust keyframe-to-keyframe motion inbetweening producing natural transitions across long time spans.

Abstract

In this work we present a novel, robust transition generation technique that can serve as a new tool for 3D animators, based on adversarial recurrent neural networks. The system synthesises high-quality motions that use temporally-sparse keyframes as animation constraints. This is reminiscent of the job of in-betweening in traditional animation pipelines, in which an animator draws motion frames between provided keyframes. We first show that a state-of-the-art motion prediction model cannot be easily converted into a robust transition generator when only adding conditioning information about future keyframes. To solve this problem, we then propose two novel additive embedding modifiers that are applied at each timestep to latent representations encoded inside the network's architecture. One modifier is a time-to-arrival embedding that allows variations of the transition length with a single model. The other is a scheduled target noise vector that allows the system to be robust to target distortions and to sample different transitions given fixed keyframes. To qualitatively evaluate our method, we present a custom MotionBuilder plugin that uses our trained model to perform in-betweening in production scenarios.

How to read this

Category
Method: a neural motion in-betweening tool
Contributions
  • An adversarial recurrent network that synthesizes transitions from temporally-sparse keyframes
  • A time-to-arrival embedding that varies transition length with a single model
  • A scheduled target-noise vector for robustness to target distortion and for sampling varied transitions from fixed keyframes
Context
Reframes the traditional keyframe in-betweening task (descended from Burtnyk-Wein keyframe animation) as learned transition generation, noting a state-of-the-art motion-prediction model does not trivially convert into a robust in-betweener.Builds on: Interactive Skeleton Techniques for Enhancing Motion Dynamics in Key Frame Animation
Correctness
Motivated by showing naive future-keyframe conditioning is insufficient, then addressed with the two embedding modifiers; demonstrated qualitatively including a MotionBuilder plugin, so robustness is shown in practice while generalization stays tied to the training motion distribution.
Clarity
Accessible narrative; the two embedding modifiers are the crux and merit a second pass.
How to read it
First pass for why plain conditioning fails and what the two embeddings fix; second pass on the time-to-arrival and scheduled-noise mechanisms, which are the transferable ideas.

Builds on

Built upon by

Related work

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 →