Semantic Animation describes using high level descriptors (such as verbal commands) to direct the performances of special character assets which are capable of determining their own detailed low-level behaviours and scene interactions. The required semantic information can be created in a manual authoring process or by machine learning. It includes our work in both virtual production and physically based animation.
- To develop and demonstrate real-time control systems for authoring animated content using smart assets, automatically synthesising new scenes from existing ones and integrating smart assets into virtual production scenarios with editable cameras and lights
- To test the prototype technologies and tool-kits in a series of experimental productions and evaluate their performance in realistic contexts of professional use.
In this context Filmakademie has crafted and released the PHS Motion Library. it contains over an hour of bipedal reference motion capture data and videos. The data-set was created with particular focus on emotional variations of walk cycles.