Kimodo: Scaling Controllable Human Motion Generation

TL;DR

Kimodo is a new kinematic motion diffusion model trained on 700 hours of mocap data, enabling high-quality, controllable human motion synthesis via text and various kinematic constraints.

Contribution

The paper introduces Kimodo, a scalable, controllable motion generation model trained on a large dataset, with a novel two-stage architecture for improved motion quality and flexibility.

Findings

Model achieves high-quality motion synthesis.

Scaling dataset and model size improves performance.

Flexible control through multiple kinematic constraints.

Abstract

High-quality human motion data is becoming increasingly important for applications in robotics, simulation, and entertainment. Recent generative models offer a potential data source, enabling human motion synthesis through intuitive inputs like text prompts or kinematic constraints on poses. However, the small scale of public mocap datasets has limited the motion quality, control accuracy, and generalization of these models. In this work, we introduce Kimodo, an expressive and controllable kinematic motion diffusion model trained on 700 hours of optical motion capture data. Our model generates high-quality motions while being easily controlled through text and a comprehensive suite of kinematic constraints including full-body keyframes, sparse joint positions/rotations, 2D waypoints, and dense 2D paths. This is enabled through a carefully designed motion representation and two-stage denoiser architecture that decomposes root and body prediction to minimize motion artifacts while allowing for flexible constraint conditioning. Experiments on the large-scale mocap dataset justify key design decisions and analyze how the scaling of dataset size and model size affect performance.