Video Motion Capture¶
Animation & Motion
Open Source¶
-
Fast SAM 3D Body Discovered by AI
2026
A training-free acceleration framework that reformulates the SAM 3D Body pipeline for real-time human mesh recovery. It replaces the iterative MHR-to-SMPL bottleneck with a direct neural mapping, achieving up to a 10.9x end-to-end speedup.
mocappose-estimationreal-time -
MoCapAnything V2 Discovered by AI
2026
An end-to-end motion capture framework that unifies video-to-pose and pose-to-rotation learning for arbitrary skeletons from monocular videos. It supports humans, animals, and custom rigged 3D assets without relying on template-specific body models.
mocappose-estimationretargetingskeleton -
Multi-HMR 2 Discovered by AI
2026
An advanced multi-person camera-centric human detection, 3D mesh recovery, and tracking framework that operates in the camera coordinate system for accurate metric 3D localization.
mocappose-estimationsmpl-x -
OnlineHMR Discovered by AI
2026
A fully online framework for video-based world-grounded human mesh recovery that operates in real-time without relying on future frames or global optimization. It is highly suitable for interactive scenarios like AR/VR, telepresence, and real-time motion capture.
human-mesh-recoverymocappose-estimation -
SAM 3D Body Discovered by AI
2026
A promptable foundation model for robust, single-image full-body 3D human mesh recovery (HMR) that estimates body, feet, and hand poses. It introduces the Momentum Human Rig (MHR) to decouple skeletal structure and surface shape, showing strong generalization in the wild.
foundation-modelhuman-mesh-recoverymocappose-estimation -
SAM3DBody-cpp Discovered by AI
2026
A high-performance, real-time C++ implementation of Meta's SAM 3D Body model. It supports multi-person 3D full-body reconstruction from a single camera, ONNX/ggml runtimes, and direct BVH motion capture file export.
bvhmocappose-estimationreal-time -
SAM4Dcap: Training-free Biomechanical Twin System from Monocular Video Discovered by AI
2026
An open-source, training-free biomechanical twin system that estimates 3D human pose and biomechanical metrics from monocular video.
biomechanicsmocapmonocularpose-estimation -
MAMMA: Markerless & Automatic Multi-Person Motion Action Capture Discovered by AI
2025
A markerless motion-capture pipeline that accurately recovers SMPL-X parameters from multi-view videos of multi-person interaction sequences.
markerlessmocapmulti-personsmpl-x -
MoCapAnything Discovered by AI
2025
A unified 3D motion capture framework that reconstructs rotation-based animations (such as BVH) for arbitrary rigged 3D assets from monocular videos, bypassing species- or template-specific limitations.
bvhmocappose-estimationskeleton -
PersPose Discovered by AI
2025
A monocular 3D human pose estimation framework that introduces Perspective Encoding and Perspective Rotation to incorporate camera intrinsics directly into cropped images. This resolves depth ambiguity and significantly improves the accuracy of 3D joint reconstruction from single-camera video.
mocappose-estimationsmpl -
PromptHMR: Promptable Human Mesh Recovery Discovered by AI
2025
An open-source framework for promptable human mesh recovery that allows incorporating auxiliary side information (such as 2D keypoints, bounding boxes, or text prompts) to improve 3D reconstruction accuracy in challenging scenarios like crowds and occlusions.
human-mesh-recoverymocappose-estimationsmpl -
SAM-Body4D Discovered by AI
2025
A training-free framework for 4D human body mesh recovery from videos. It leverages temporal continuity and diffusion-based video object segmentation to extend SAM 3D Body to video sequences, ensuring temporal consistency and robustness under occlusions.
4d-reconstructionmocappose-estimationvideo-mocap -
SMPLest-X: Ultimate Scaling for Expressive Human Pose and Shape Estimation Discovered by AI
2025
An extended and simplified version of SMPLer-X that scales up expressive human pose and shape estimation. It replaces the component guiding module with a minimalist transformer decoder, significantly boosting hand and body pose estimation accuracy.
mocappose-estimationsmpl-xGitHub · arXiv · Hugging Face
-
GVHMR: World-Grounded Human Motion Recovery via Gravity-View Coordinates Discovered by AI
2024
A world-grounded human motion recovery framework that estimates gravity-aware human poses from monocular video. By defining a novel Gravity-View coordinate system, it leverages gravity priors to avoid accumulated errors and reconstructs stable 3D motion in global coordinates.
mocappose-estimationsmpl -
Multi-HMR Discovered by AI
2024
A single-shot model for multi-person whole-body 3D human mesh recovery from a single RGB image, predicting body, hands, facial expressions, and 3D location in camera coordinates.
human-mesh-recoverymocappose-estimation -
Sapiens Discovered by AI
2024
A family of high-resolution foundation models developed by Meta for human-centric vision tasks, including 2D/3D pose estimation, body-part segmentation, and depth estimation.
foundation-modelmocappose-estimation -
TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representation Discovered by AI
2024
Introduces a discrete token-based pose representation and a Threshold-Adaptive Loss Scaling (TALS) loss to constrain predictions to valid human poses, achieving state-of-the-art accuracy in 3D human mesh recovery from single images and video frames.
mesh-recoverymocappose-estimationsmpl -
TRAM: Global Trajectory and Motion of 3D Humans from in-the-wild Videos Discovered by AI
2024
A two-stage method to reconstruct a human's global trajectory and motion from in-the-wild videos by robustifying SLAM to recover camera motion and using a video transformer model to regress kinematic body motion.
mocappose-estimationsmpltrajectory -
4D-Humans (HMR 2.0) Discovered by AI
2023
A fully transformer-based framework for reconstructing and tracking human meshes over time from monocular video, advancing the state of the art in handling complex poses.
mesh-recoverymocaptrackingtransformer -
Co-Evolution of Pose and Mesh for 3D Human Body Estimation from Video Discovered by AI
2023
A video-based 3D human mesh recovery framework that introduces 3D pose as an intermediary and utilizes a co-evolution decoder to perform pose and mesh interactions for consistent and smooth motion estimation.
mesh-recoverymocappose-estimationsmpl -
HaMeR Discovered by AI
2023
A robust transformer-based hand mesh recovery model designed to reconstruct 3D hands from monocular input with high accuracy and robustness.
hand-posemesh-recoverymocaptransformer -
HybrIK-X: Hybrid Analytical-Neural Inverse Kinematics for Whole-Body Mesh Recovery Discovered by AI
2023
A holistic whole-body mesh recovery framework that combines 3D keypoint estimation with a hybrid analytical-neural inverse kinematics (IK) solver to reconstruct expressive body, hand, and face meshes.
inverse-kinematicsmesh-recoverymocappose-estimation -
PoseFormerV2 Discovered by AI
2023
An efficient and robust transformer-based 3D human pose estimation framework that operates in the frequency domain. It reduces computational complexity and improves robustness against noisy 2D joint detections in long video sequences.
3d-posepose-estimationtransformer -
SLAHMR: Simultaneous Localization and Human Mesh Recovery Discovered by AI
2023
An optimization-based framework that decouples human and camera motion from monocular videos in the wild. It places multiple moving people and the camera into a unified global coordinate frame, enabling accurate trajectory tracking and motion capture.
mocappose-estimationtracking -
SMPLer-X Discovered by AI
2023
A generalist foundation model for expressive human pose and shape estimation, scaling up training to 4.5M instances to capture body, hands, and face motion.
foundation-modelmocappose-estimationsmpl-x -
WHAM Discovered by AI
2023
A framework for reconstructing world-grounded 3D human motion from monocular video, integrating camera motion estimation and temporal context to achieve accurate global trajectories and reduce foot sliding.
human-mesh-recoverymocappose-estimation -
CLIFF: Carrying Location Information in Full Frames into Human Pose and Shape Estimation Discovered by AI
2022
A top-down human pose and shape estimation method that carries full-frame location information into the crop-based network. This global perspective enables highly accurate prediction of global rotation and translation in the original camera coordinate system.
mocappose-estimationsmpl -
DeciWatch: A Simple Baseline for 10x Efficient 2D and 3D Pose Estimation Discovered by AI
2022
A highly efficient framework for video-based 2D and 3D human pose estimation that achieves a 10x speedup by sparsely sampling video frames and using a sample-denoise-recover pipeline to reconstruct the full motion sequence.
efficiencymocappose-estimationtemporal -
MotionBERT Discovered by AI
2022
A unified pre-training framework that learns human motion representations from heterogeneous data to solve 3D pose estimation, mesh recovery, and action recognition.
mocapmotion-representationpose-estimationpre-training -
XR Animator Discovered by AI
2022
An open-source, AI-based full-body motion capture application designed for VTubing and 3D character animation, supporting real-time tracking of body, face, and hands from any standard webcam or video file.
mocapvtuberwebcam -
BlendArMocap Discovered by AI
2021
A Blender add-on that enables real-time markerless motion tracking and capture directly inside Blender using MediaPipe and Rigify.
blendermediapipemocap -
EasyMocap Discovered by AI
2021
An open-source toolbox for markerless human motion capture and 3D reconstruction from monocular or multi-view videos.
3d-reconstructionmarkerlessmocapmulti-view -
FreeMoCap Discovered by AI
2021
An open-source, markerless 3D motion capture system that allows users to extract high-quality motion data from video recordings using consumer-grade cameras.
3d-motionmarkerlessmocappose-estimation -
GLAMR: Global Occlusion-Aware Human Mesh Recovery with Dynamic Cameras Discovered by AI
2021
A global human mesh recovery method designed for monocular videos captured by dynamic cameras. It integrates a deep generative motion infiller to reconstruct continuous human trajectories and poses even during severe or long-term occlusions.
mocapocclusionpose-estimation -
SysMocap Discovered by AI
2021
A real-time 3D motion capture system designed for virtual character animation and VTubing, utilizing lightweight pose estimation models.
mocapreal-timevtuber -
FrankMocap Discovered by AI
2020
An open-source, fast, and easy-to-use system for estimating 3D hand and body poses simultaneously from single-view videos.
3d-posebody-posehand-posemocap -
MMPose Discovered by AI
2020
OpenMMLab's comprehensive pose estimation toolbox providing state-of-the-art models for 2D/3D human pose, whole-body tracking, and video motion capture.
3d-posemocappose-estimation -
VIBE: Video Inference for Human Body Pose and Shape Estimation Discovered by AI
2019
A video-based 3D human pose and shape estimation method that leverages a recurrent neural network (GRU) and adversarial training with the AMASS motion capture dataset to produce temporally consistent and realistic SMPL body model sequences.
mesh-recoverymocappose-estimationsmpl -
VideoPose3D Discovered by AI
2018
An efficient 3D human pose estimation model in video using dilated temporal convolutions over 2D keypoint trajectories. It introduces a semi-supervised training method using back-projection to leverage unlabeled video data.
3d-posepose-estimationtemporal-convolution
Proprietary / Industry Tools¶
-
Rokoko Vision Discovered by AI
2023
A browser-based AI motion capture tool that extracts 3D character animations from webcam feeds or uploaded video files. It supports single-camera and dual-camera setups for enhanced tracking accuracy and integrates with Rokoko Studio for cleanup and retargeting.
browser-baseddual-cameramocappose-estimation -
Wonder Studio Discovered by AI
2023
An AI-powered cloud platform that automatically animates, lights, and composes 3D CG characters into live-action video. It performs multi-camera or single-camera motion capture, camera tracking, and actor replacement, exporting clean FBX/OBJ motion data and camera tracks to Blender, Maya, or Unreal Engine.
actor-replacementmocapvfx -
Rokoko Video Discovered by AI
2022
A free browser-based AI motion capture tool by Rokoko that extracts 3D character animation from video clips or webcams, exporting directly to FBX, BVH, or major game engines.
animationbrowser-basedfreemocap -
Kinetix Discovered by AI
2020
An AI-powered 3D human motion platform that converts video clips into custom 3D character animations and emotes. It automates motion capture, cleaning, and retargeting, allowing creators to apply animations to any compatible 3D avatar and stream them directly into games or virtual worlds.
emotesmocapretargeting -
A commercial markerless motion-capture service that converts ordinary video into 3D animation (FBX/BVH/GLB), enabling mocap without a suit or studio.
apicommercialmarkerless-mocap -
A markerless multi-camera motion-capture service producing high-quality 3D motion from standard cameras, used by indie and mid-size studios.
apicommercialmarkerless-mocap
.jpg)
