I want to introduce my motion matching system for ue4. The peculiarity of this system is that not used large mocap sessions (dance cards). That works with classic game animations (locomotion cycles, idle poses etc.).
How it works?
0. Register first all assets (animations/blendspaces) who will participate in matching system. Today its realized in special blueprint, but ideally its gonna be implemented in anim graph.
- We have a standart anim graph logic with blend trees, blendspaces, state machines and other…
- All matched sequences/blendspaces converted from “asset player” to “one frame animation” node type (for access to manually control animation time)
- Every frame system detect leader registered asset (who have maximum weight in anim graph), and fallower assets. Each asset have leader/fallower behavior, defined on register stage.
Matching algoritms are similar to other motion matching systems: pose/predicted pose/velocity matching, and special matching algorithms for idle state.
At the moment only the core has been created. Many features like movement stuttering fix or continious interpolation from leader assets implemented in other test projects. The main task now is to translate everything into code, thats allows registed assets directly in anim graph, improve perfomance and usability.
Now i need technical help from animation programmers on intergation whole system in anim graph.
Sorry for my English.
UPDATE: matching system was totally reworked. Now all animations syced with default sync groups.