Kimodo: Scaling Controllable Human Motion Generation

2026-03-16

AI Summary

This work proposes Kimodo, a controllable human motion diffusion model trained on 700 hours of optical motion capture data, addressing the limitations imposed by the small scale of existing public motion datasets, which hinder the performance of generative models in terms of motion quality, control accuracy, and generalization. Kimodo employs a two-stage denoising architecture that separately models root trajectory and body pose, complemented by a dedicated motion representation designed to reduce artifacts. The framework flexibly integrates textual instructions with diverse kinematic constraints—including keyframes, joint poses, and 2D paths—enabling precise and expressive motion synthesis. Experimental results demonstrate that the combination of large-scale training data and the proposed architectural design significantly enhances both the quality and controllability of generated motions.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: Motion and Path Planning
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

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.

Problem

• human motion generation
• motion capture dataset
• controllable generation
• data scaling
• motion quality

Innovation

• kinematic motion diffusion
• controllable motion generation
• large-scale mocap
• two-stage denoiser
• multi-modal constraints