Data captured with motion capture systems almost always involve retargeting, which is the process of taking captured motion and applying it to characters with different proportions or skeletal structures. This is still a process mostly done by hand. Using neural networks, RetargetX can more accurately predict how a movement would look on varied character models, taking physics into account, leading to more natural and realistic animations and minimizing the amount of manual work to do it.
The power of neural networks allows RetargetX to adapt motion data not only to characters with different builds and proportions, but also motion styles, ensuring that every movement remains fluid, lifelike, and true to the model’s motion characteristics regardless of the character’s design. For example, RetargetX could convert a performer’s gait cycle to a completely different gait style that has been predefined by the character’s motion design team.