We leverage a pretrained diffusion UNet backbone for controlled human image animation, enabling expressive dynamic details and precise motion control. Specifically, we introduce a dynamics adapter ...