From Keyframes to Directing Performances: Redefining Character Animation with AI

Although the character-animation process has been largely digitized, the fundamental approach has remained relatively unchanged for more than a century. Animators create performances by creating key poses, which they then string together in a sequence to create motion. They continue to add and finesse these keyframes until they get a believable performance. In this presentation, we will demonstrate a new approach to character animation that uses AI to transform the way animators work. With neural motion controllers, we use machine learning to model how characters move and behave. An animator then only needs to specify high-level intent, like when and where they want the character to move, and the neural motion controller does the tedious work of producing natural, believable motion. We will also demonstrate that by connecting this new technique with core Autodesk capabilities, we can create a production-ready workflow where animators retain full artistic control over the final animation.