🏃‍♂️Text-to-Animation

Spatial Scene & Animation (Text-to-Animation Systems): We have been able to isolate systems that generate animations from natural language input, focusing on those that perform some natural language understanding and are not purely data-driven.

We provide a text-to-animation solution that allows users to edit 3D animation by simply writing and editing texts. Our design objectives are multifold: first, we aim to generate virtual scenes close to the daily life where objects can be interacted with; second, we expect the character to be aware of the scene context and adaptively locomote in the scene; third, diverse character motions should be supported, such as waving arm angrily, playing the piano, etc. To accomplish the objectives, we build our system upon Language-Driven Synthesis [MPF∗18] and NSM-based character locomotion [SZKS19] to jointly optimize the scene layout and character motion, and further generate vivid 3D animation by editing texts.