HumanCM: One Step Human Motion Prediction

Abstract--We present HumanCM, a one-step human motion prediction framework built upon consistency models. Instead of relying on multi-step denoising as in diffusion-based methods, HumanCM performs efficient single-step generation by learning a self-consistent mapping between noisy and clean motion states in a latent space. By operating in this compact representation, HumanCM captures long-range temporal dependencies and preserves motion coherence. Experiments on Human3.6M and HumanEva-I demonstrate that HumanCM achieves comparable or superior accuracy to state-of-the-art diffusion models while reducing inference steps by up to two orders of magnitude. Human motion prediction (HMP) is a fundamental task in computer vision and robotics, aiming to forecast future 3D human poses from observed motion sequences.