| Overview: |
This presentation will introduce to you a character animation system using machine learning, dubbed MotorNerve, which implements two core character animation functions: locomotion control and transition animation generation. First, by combining Motion Matching and Learned Motion Matching technologies, MotorNerve achieves high-quality, low-consumption locomotion control while speeding up the tuning process for Learned Motion Matching. Secondly, MotorNerve also implements an inhouse Motion In-Betweening (MIB) algorithm, which uses a variational autoencoder to encode leg movements, significantly improving the foot skating problem of existing methods. This method, a joint work by Zhejiang University and TiMi Studio Group, a Tencent Games studio, has been published in SIGGRAPH 2022. MotorNerve applies MIB to interactive animation scenarios, generating high-quality transition animations while saving the project costs.
|
