Motion Retargeting¶
Animation & Motion
Open Source¶
-
AnyAct: Towards Human Reenactment of Character Motion From Video Discovered by AI
2026
A framework that formulates character-video-driven human reenactment as conditional human motion generation from transferable sparse local 2D articulated motion, reinterpreting non-human character motion as plausible human performances.
character-animationmotion-retargetingvideo-to-motion -
Make Tracking Easy: Neural Motion Retargeting for Humanoid Whole-body Control Discovered by AI
2026
A neural motion retargeting (NMR) framework that learns a direct mapping from human SMPL motion sequences to physically plausible humanoid robot motions, bypassing local-optima failures of conventional optimization-based methods.
humanoidmotion-retargetingneural-networkwhole-body-control -
ReActor Discovered by AI
2026
A physics-aware motion retargeting framework developed by Disney Research that uses reinforcement learning and bilevel optimization to adapt human reference motions to diverse robotic and character morphologies without physical artifacts.
motion-retargetingphysics-awarereinforcement-learning -
Skinned Motion Retargeting via Artifact-driven Kinematic Prior Refinement Discovered by AI
2026
A geometry-aware motion retargeting framework that addresses self-penetration artifacts on the posed target mesh by converting them into corrective cues in a shared motion embedding via a motion-to-vertex Jacobian.
geometry-awareretargetingskinned-character -
Skinned Motion Retargeting with Spatially Adaptive Interaction Guidance Discovered by AI
2026
A geometry-aware motion retargeting framework that preserves interaction semantics, such as self-contact and near-body proximity, across characters with varying body shapes by performing proximity matching over dynamically adapted anchors.
geometry-awareinteractionretargetingskinned-character -
SOMA: Unifying Parametric Human Body Models Discovered by AI
2026
A unified parametric human body model and retargeting library (soma-retargeter) that converts SOMA human motion captures into humanoid robot joint animations using GPU-optimized inverse kinematics via Newton and NVIDIA Warp.
bvhhumanoidikretargeting -
Unified Motion Retargeting for Humanoids with Learned Point Cloud Correspondence Discovered by AI
2026
A framework that learns dense point cloud correspondence between human and humanoid robot surfaces without requiring manually designed keypoint mappings. It optimizes robot motion using matched surface positions, orientations, contacts, and kinematic constraints.
humanoidmotion-retargetingpoint-cloudrobotics -
GMR: General Motion Retargeting Discovered by AI
2025
A general motion retargeting framework (GMR) that maps human motion capture data to diverse humanoid robots in real time on CPU, preserving key physical and kinematic constraints.
humanoidmocapretargetingrobotics -
IKMR: Implicit Kinodynamic Motion Retargeting Discovered by AI
2025
A scalable and efficient motion retargeting framework designed for human-to-humanoid imitation learning that considers both kinematics and dynamics to produce smooth, physically feasible, and jitter-free robot trajectories from noisy human motion data.
humanoidimitation-learningretargetingrobotics -
MimiCAT Discovered by AI
2025
A cascade-transformer framework designed for category-free 3D pose transfer that learns soft correspondences across keypoints of structurally diverse characters (e.g., humanoids to quadrupeds).
3d-animationmotion-retargetingpose-transfer -
MoReFlow Discovered by AI
2025
An unsupervised motion retargeting framework via Flow Matching that learns correspondences between characters' motion embedding spaces using a VQ-VAE tokenizer, enabling flexible and reversible retargeting without paired data.
flow-matchingretargetingunsupervised -
OmniRetarget Discovered by AI
2025
OmniRetarget is an interaction-preserving motion retargeting framework designed for humanoid whole-body loco-manipulation and scene interaction. It addresses the embodiment gap by optimizing for physical plausibility, preventing artifacts like foot-skating and self-penetration while preserving human-object and human-environment interactions.
humanoidloco-manipulationphysics-awareretargeting -
PALUM: Part-based Attention Learning for Unified Motion Retargeting Discovered by AI
2025
PALUM presents a part-based attention learning framework for motion retargeting across distinct skeleton topologies by partitioning joints into semantic body parts and learning common motion representations.
attentioncharacter-animationmotion-retargetingskeleton -
ReConForM: Real-time Contact-aware Motion Retargeting for more Diverse Character Morphologies Discovered by AI
2025
ReConForM is a real-time contact-aware motion retargeting framework that automatically transfers rigged key vertices to preserve self-collisions and ground contacts across diverse character morphologies.
contact-awaremotion-retargetingreal-time -
STaR: Seamless Spatial-Temporal Aware Motion Retargeting with Penetration and Consistency Constraints Discovered by AI
2025
STaR is a spatial-temporal aware motion retargeting framework that incorporates penetration and consistency constraints to achieve a balance between motion semantics preservation, physical plausibility, and temporal consistency.
character-animationmotion-retargetingtemporal-consistency -
BVH Motion Retargeter Discovered by AI
2024
A Blender add-on designed to convert and retarget BVH motion-capture data to common humanoid skeleton conventions, including Mixamo, Unreal Engine 5, and VRM humanoid avatars.
blenderbvhmixamoue5 -
GMR: General Motion Retargeting Discovered by AI
2024
General Motion Retargeting (GMR) is a real-time, CPU-executable motion retargeting library designed to map human motion capture data onto diverse humanoid robot and character skeletal structures.
character-animationhumanoidmotion-retargetingreal-time -
MeshRet: Skinned Motion Retargeting with Dense Geometric Interaction Perception Discovered by AI
2024
MeshRet is a skinned motion retargeting framework that captures and aligns dense geometric interaction fields between body parts to prevent self-interpenetration and preserve contact semantics.
character-animationmesh-retargetingmotion-retargeting -
robot_retargeter Discovered by AI
2024
A toolkit for retargeting human motion (SMPL-X) or source-robot motion to target humanoid robots, featuring multi-robot side-by-side visualization and inverse kinematics via mink and MuJoCo.
humanoidikmujocosmpl-x -
SOMA Retargeter Discovered by AI
2024
A GPU-optimized motion retargeting library built on NVIDIA Warp and Newton, designed to convert human motion capture data (BVH files) into humanoid robot joint animations with feet stabilization and joint limit constraints.
humanoidmocapretargetingwarp -
Towards High-Quality 3D Motion Transfer with Realistic Apparel Animation Discovered by AI
2024
Presents MMDMC, a 3D motion transfer framework that models physical dynamics and local deformations of clothing separately from the rigid body skeleton. This approach prevents body-apparel penetration and produces highly detailed, realistic apparel animations for stylized characters.
3d-characterapparel-animationeccvmotion-transfer -
Perpetual Humanoid Control for Real-time Simulated Avatars Discovered by AI
2023
A physics-based character animation framework that trains a single robust policy (Perpetual Humanoid Controller) to perform real-time motion imitation and retargeting across diverse body shapes and proportions.
avatarimitationphysicsretargeting -
Physics-based Motion Retargeting from Sparse Inputs Discovered by AI
2023
A physics-based motion retargeting framework that uses reinforcement learning to train a torque-control policy, enabling real-time motion retargeting from sparse human sensor data to characters of various morphologies in a physics simulator.
physicsreinforcement-learningretargeting -
Pose-aware Attention Network (PAN) Discovered by AI
2023
A flexible motion retargeting framework that decomposes skeletons into body parts and utilizes a pose-aware attention network to transfer motion between characters with different structures, such as humans and quadrupeds.
attentioncharacter-animationcross-structureretargeting -
Pose-to-Motion: Cross-Domain Motion Retargeting with Pose Prior Discovered by AI
2023
A novel motion-retargeting approach for motion synthesis that transfers motion from an existing motion capture dataset of a source character to a target character using only static pose data from the target domain. This addresses the scarcity of high-quality motion data for diverse character skeletons.
motion-retargetingmotion-synthesispose-prior -
R2ET: Skinned Motion Retargeting with Residual Perception of Motion Semantics & Geometry Discovered by AI
2023
R2ET (Residual RETargeting network) is a skinned motion retargeting framework that utilizes a skeleton-aware module to preserve motion semantics and a shape-aware module to perceive and adjust for target character geometry.
character-animationmotion-retargetingskinned-mesh -
Semantics-aware Motion Retargeting with Vision-Language Models Discovered by AI
2023
A semantics-aware motion retargeting (SMT) method that leverages vision-language models to extract and maintain high-level motion semantics, ensuring meaningful motion transfer between characters.
character-animationretargetingsemanticsvlm -
Skeleton-Aware Networks for Deep Motion Retargeting Discovered by AI
2020
A deep learning framework for unsupervised motion retargeting between skeletons of different structures by leveraging skeletal pooling and unpooling operations to reduce homeomorphic graphs to a common primal skeleton.
motion-retargetingskeleton-awareunsupervised-learning -
TransMoMo Discovered by AI
2020
An unsupervised video motion retargeting method that transfers motion between different subjects in monocular videos by decomposing motion, structure, and view-angle in an invariance-driven latent space.
motion-transferunsupervisedvideo-retargeting
Proprietary / Industry Tools¶
-
Autodesk MotionBuilder Discovered by AI
2026
An industry-standard 3D character animation software used for virtual cinematography, motion capture editing, and real-time motion retargeting onto complex character rigs.
animation-softwareindustry-standardmocapretargeting -
Plask Discovered by AI
2024
A browser-based AI-powered motion capture and animation platform that allows users to extract 3D motion from video and retarget it to custom characters.
ai-animationbrowser-basedmocapretargeting -
Unreal Engine IK Retargeter Discovered by AI
2022
A built-in animation retargeting tool in Unreal Engine 5 that allows developers to transfer animations between characters with different skeletal hierarchies and proportions using IK Rigs.
animation-transferik-rigunreal-engine -
Rokoko Studio Discovered by AI
2020
A popular commercial motion capture and retargeting software suite that allows creators to record, visualize, and retarget full-body, facial, and finger animations to custom 3D characters with real-time integration for Blender, Maya, and Unreal Engine.
blender-plugincommercialmocapretargeting -
Reallusion iClone Discovered by AI
2005
A professional real-time 3D animation and character creation software that features advanced motion retargeting, motion editing, and live motion capture integration for game development and film production.
animationcharacter-creationcommercialretargeting
