A Self-Supervised Approach on Motion Calibration for Enhancing Physical Plausibility in Text-to-Motion

Generating semantically aligned human motion from textual descriptions has made rapid progress, but ensuring both semantic and physical realism in motion remains a challenge. In this paper, we introduce the Distortion-aware Motion Calibrator (DMC), a post-hoc module that refines physically implausible motions (e.g., foot floating) while preserving semantic consistency with the original textual description. Rather than relying on complex physical modeling, we propose a self-supervised and data-driven approach, whereby DMC learns to obtain physically plausible motions when an intentionally distorted motion and the original textual descriptions are given as inputs. We evaluate DMC as a post-hoc module to improve motions obtained from various text-to-motion generation models and demonstrate its effectiveness in improving physical plausibility while enhancing semantic consistency. The experimental results show that DMC reduces FID score by 42.74% on T2M and 13.20% on T2M-GPT, while also achieving the highest R-Precision. When applied to high-quality models like MoMask, DMC improves the physical plausibility of motions by reducing penetration by 33.0% as well as adjusting floating artifacts closer to the ground-truth reference. These results highlight that DMC can serve as a promising post-hoc motion refinement framework for any kind of text-to-motion models by incorporating textual semantics and physical plausibility.

• Publication year

  2026

• Venue

  Unknown venue

• Publication date

  2026-02-20

• Fields of study

  Computer Science

• Identifiers
  arXiv 2602.18199
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Humanoid Locomotion and Manipulation: Current Progress and Challenges in Control, Planning, and Learning

  2025cited by this paper
• Sketch2Anim: Towards Transferring Sketch Storyboards into 3D Animation

  2025cited by this paper
• FlexMotion: Lightweight, Physics-Aware, and Controllable Human Motion Generation

  2025cited by this paper
• DartControl: A Diffusion-Based Autoregressive Motion Model for Real-Time Text-Driven Motion Control

  2024cited by this paper
• CLoSD: Closing the Loop between Simulation and Diffusion for multi-task character control

  2024cited by this paper
• OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learning

  2024cited by this paper
• Toward User-Aware Interactive Virtual Agents: Generative Multi-Modal Agent Behaviors in VR

  2024cited by this paper
• Harmon: Whole-Body Motion Generation of Humanoid Robots from Language Descriptions

  2024cited by this paper
• ReinDiffuse: Crafting Physically Plausible Motions with Reinforced Diffusion Model

  2024influential reference
• ReMoDiffuse: Retrieval-Augmented Motion Diffusion Model

  2023cited by this paper
• MoMask: Generative Masked Modeling of 3D Human Motions

  2023influential reference
• ImitationNet: Unsupervised Human-to-Robot Motion Retargeting via Shared Latent Space

  2023cited by this paper
• MotionGPT: Finetuned LLMs are General-Purpose Motion Generators

  2023cited by this paper
• Guided Motion Diffusion for Controllable Human Motion Synthesis

  2023cited by this paper
• Make-An-Animation: Large-Scale Text-conditional 3D Human Motion Generation

  2023cited by this paper
• TM2D: Bimodality Driven 3D Dance Generation via Music-Text Integration

  2023cited by this paper
• Generating Human Motion from Textual Descriptions with Discrete Representations

  2023influential reference
• Generating Diverse and Natural 3D Human Motions from Text

  2022cited by this paper
• TEMOS: Generating diverse human motions from textual descriptions

  2022cited by this paper
• MotionMixer: MLP-based 3D Human Body Pose Forecasting

  2022cited by this paper
• MotionDiffuse: Text-Driven Human Motion Generation With Diffusion Model

  2022cited by this paper
• Human Motion Diffusion Model

  2022cited by this paper
• Learning Variational Motion Prior for Video-based Motion Capture

  2022cited by this paper
• PhysDiff: Physics-Guided Human Motion Diffusion Model

  2022influential reference
• Learning Transferable Visual Models From Natural Language Supervision

  2021cited by this paper
• Skeletor: Skeletal Transformers for Robust Body-Pose Estimation

  2021cited by this paper
• SmoothNet: A Plug-and-Play Network for Refining Human Poses in Videos

  2021cited by this paper
• Investigating Pose Representations and Motion Contexts Modeling for 3D Motion Prediction

  2021cited by this paper
• Text2Gestures: A Transformer-Based Network for Generating Emotive Body Gestures for Virtual Agents**This work has been supported in part by ARO Grants W911NF1910069 and W911NF1910315, and Intel. Code and additional materials available at: https://gamma.umd.edu/t2g

  2021cited by this paper
• Denoising Diffusion Probabilistic Models

  2020cited by this paper
• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

  2020cited by this paper
• A scalable approach to control diverse behaviors for physically simulated characters

  2020cited by this paper
• Learning Trajectory Dependencies for Human Motion Prediction

  2019cited by this paper
• Learning Diverse Stochastic Human-Action Generators by Learning Smooth Latent Transitions

  2019cited by this paper
• Learning predict-and-simulate policies from unorganized human motion data

  2019cited by this paper
• Learning basketball dribbling skills using trajectory optimization and deep reinforcement learning

  2018cited by this paper
• Learning to Schedule Control Fragments for Physics-Based Characters Using Deep Q-Learning

  2017cited by this paper
• Improved Training of Wasserstein GANs

  2017cited by this paper
• Text2Action: Generative Adversarial Synthesis from Language to Action

  2017cited by this paper
• Learning human behaviors from motion capture by adversarial imitation

  2017cited by this paper
• Deep Video Generation, Prediction and Completion of Human Action Sequences

  2017cited by this paper
• A Deep Learning Framework for Character Motion Synthesis and Editing

  2016cited by this paper
• Recurrent Network Models for Human Dynamics

  2015cited by this paper
• Simulating humans: computer graphics animation and control

  1993cited by this paper

• No citing papers are available for this paper.