This dataset contains approximately 199 hours and 30 minutes of full-length continuous motion capture data, including Motive BVH, NVIDIA SOMA Uniform BVH, Unitree G1 CSV, and structured metadata.

For commercial licensing or other enquiries, contact us at hello@appleartsstudios.com.

By submitting this form, you confirm that the information provided is accurate and agree to the dataset terms.

Apple Arts Studios

Apple Arts Studios Full-Length Motion Capture Dataset

Apple Arts Studios Full-Length Motion Capture Dataset

Dataset Overview

The Apple Arts Studios Full-Length Motion Capture Dataset is a professionally captured, full-body human-motion dataset containing approximately 199 hours and 30 minutes of continuous motion capture data.

Unlike segmented motion datasets, this repository preserves the complete capture sequences without separating individual actions into short clips. The recordings retain the original capture timeline and may include:

T-pose calibration Idle motion Preparation movements Transitions between actions Performer repositioning Repeated motion variations Settle time between performances Complete performed motion sequences

The dataset provides multiple motion representations for animation, AI, and humanoid robotics workflows:

OptiTrack Motive Proportional BVH NVIDIA SOMA UNIFORM BVH Unitree G1 motion CSV Structured metadata

The Unitree G1 CSV data provides robot-oriented motion representations derived from the captured human performances, enabling the dataset to support humanoid robotics, motion retargeting, imitation learning, and Physical AI research.

The accompanying metadata provides structured information associated with the motion takes and their corresponding exported representations.

The dataset is designed to support research and development in areas including:

Artificial Intelligence Humanoid Robotics Physical AI Imitation Learning Human Motion Modelling Motion Retargeting Whole-Body Motion Research Robotics Simulation Character Animation Action and Motion Understanding

Full-length captured motion: approximately 199 hours 30 minutes

The data in this repository represents continuous capture sequences and is intentionally not segmented into individual action clips.

Dataset Origin

This dataset was conceived, performed, captured, processed, and curated entirely in-house at Apple Arts Studios as part of our internal motion capture, AI, and robotics dataset development program.

All human motion performances were recorded at our own professional motion capture facility using an OptiTrack optical motion capture system.

The captured performances were subsequently processed and exported into multiple representations, including:

OptiTrack Motive Proportional BVH
NVIDIA SOMA UNIFORM BVH
Unitree G1 motion CSV
Structured metadata

The dataset preserves the full-length continuous capture takes, including calibration poses, transitions, idle periods, repositioning, and complete performed motion sequences.

Dataset Statistics

Dataset Property	Details
Total Motion Duration	Approximately 199 hours 30 minutes
Data Type	Full-length continuous motion capture
Segmentation	Non-segmented
Motion Content	T-pose, idle motion, transitions, repositioning, performed actions, and complete capture sequences
Human Motion Format	OptiTrack Motive Proportional BVH
Standardized Motion Format	NVIDIA SOMA UNIFORM BVH
Humanoid Robot Format	Unitree G1 CSV
Metadata	Structured metadata accompanying the motion data
Primary Applications	Humanoid robotics, Physical AI, imitation learning, motion retargeting, simulation, AI research, and animation

File Formats & Deliverables

Each full-length capture take is provided in multiple representations to support different animation, AI, and robotics workflows.

OptiTrack Motive Proportional BVH

The Motive Proportional BVH preserves the human skeleton proportions and full-body motion exported from the OptiTrack Motive processing workflow.

Format: .bvh

Suitable for:

Human motion analysis
Character animation
Motion retargeting
DCC workflows
Research applications

NVIDIA SOMA UNIFORM BVH

The SOMA UNIFORM BVH provides a standardized skeleton representation of the captured motion for machine-learning, motion-processing, and research workflows.

Format: .bvh

Suitable for:

Human motion modelling
Motion learning
AI research
Motion retargeting
Dataset standardization

Unitree G1 Motion Data

Human motion is additionally represented for the Unitree G1 humanoid robot.

Format: .csv

This representation is intended to support workflows involving:

Humanoid robotics
Human-to-robot motion retargeting
Imitation learning
Robotics simulation
Whole-body control research
Physical AI

Structured Metadata

Structured metadata is provided alongside the motion files to associate individual capture takes with their corresponding motion representations and dataset information.

The metadata is intended to make the dataset easier to organize, search, process, and integrate into automated AI and robotics pipelines.

Full-Length Take Structure

This dataset preserves the complete motion capture take rather than extracting only the main performed action.

Depending on the recording, a full-length take may contain:

T-pose / calibration pose
Initial idle or standing motion
Preparation before the performance
Main performed action or action sequence
Transitions between movements
Repeated variations of an action
Natural performer repositioning
Recovery or reset movements
End-of-take idle / settle time

These elements are intentionally retained so researchers and developers can work with the continuous temporal structure of the original motion capture performance.

Non-Segmented Data

The motions in this repository are not divided into individual short action clips.

For example, a single continuous take may contain:

T-Pose → Idle → Preparation → Action → Transition → Action Variation → Repositioning → Idle

This differs from the segmented Apple Arts Studios motion dataset, where individual useful motion sequences are extracted and provided as separate clips.

The full-length format can be useful for research involving:

Temporal motion understanding
Action segmentation
Motion transition modelling
Continuous behaviour learning
Sequence modelling
Automatic motion annotation
Human activity recognition
Long-horizon imitation learning
Humanoid motion generation

Dataset Directory Structure

The dataset is organized around full-length motion capture takes. Corresponding motion representations and metadata are linked using consistent take identifiers.

A typical dataset structure is:

Full_Length_Motion_Capture_Dataset/
│
├── Motive_BVH/
│   ├── Take_0001.bvh
│   ├── Take_0002.bvh
│   └── ...
│
├── SOMA_BVH/
│   ├── Take_0001.bvh
│   ├── Take_0002.bvh
│   └── ...
│
├── Unitree_G1_CSV/
│   ├── Take_0001.csv
│   ├── Take_0002.csv
│   └── ...
│
└── Metadata/
    ├── dataset_metadata.csv
    └── ...

The same take identifier is used across the available representations so users can associate the Motive BVH, SOMA BVH, Unitree G1 CSV, and metadata belonging to the same captured performance.

The motion files remain full-length and continuous and are not reorganized into individual action categories or short segmented clips.

Capture & Processing Information

All motion data in this dataset was captured at Apple Arts Studios using a professional OptiTrack optical motion capture system.

The capture and processing workflow includes:

Optical full-body motion capture
Marker tracking and reconstruction
Skeleton solving and motion processing in OptiTrack Motive
Proportional Motive BVH export
Standardization to NVIDIA SOMA UNIFORM BVH
Human-to-Unitree G1 motion conversion
Unitree G1 CSV generation
Metadata generation and validation
Final full-length dataset packaging

Unlike the segmented version of the Apple Arts Studios Motion Dataset, the source capture sequences in this repository are retained as continuous full-length takes.

No action-based trimming or segmentation is applied to the released full-length sequences. Calibration poses, idle periods, transitions, repositioning, and other naturally occurring motion between performed actions may therefore remain in the files.

Motion Representations

The same captured human performance is processed into multiple representations:

Human Capture → Motive Proportional BVH → SOMA UNIFORM BVH → Unitree G1 CSV

This allows the dataset to support both human-motion research and humanoid-robotics workflows while maintaining correspondence between the different motion representations.

Metadata and File Naming

Every motion file follows a descriptive and consistent naming convention that identifies the captured motion directly from the filename.

The dataset does not use generic sequential filenames such as:

Take_0001, Take_0002, Take_0003

Instead, filenames are based on the actual motion or activity performed during the capture.

Corresponding motion representations use consistent naming so that the same performance can be matched across:

OptiTrack Motive Proportional BVH
NVIDIA SOMA UNIFORM BVH
Unitree G1 CSV
Structured metadata

Structured Metadata

Metadata is provided separately to describe and organize the available motion data.

The metadata is intended to support:

Dataset indexing
File identification
Motion organization
Cross-format correspondence
Automated dataset processing
AI and robotics data pipelines

Intended Uses

The Apple Arts Studios Full-Length Motion Capture Dataset is designed for research, development, and experimentation involving continuous human motion and humanoid robotics.

Potential applications include:

Humanoid Robotics
Physical AI
Imitation Learning
Reinforcement Learning
Human-to-Robot Motion Retargeting
Whole-Body Control Research
Human Motion Modelling
Motion Generation
Motion Prediction
Continuous Motion Understanding
Action Segmentation Research
Sequence Modelling
Robotics Simulation
Character Animation
Motion Retargeting
Embodied AI Research

Because the dataset preserves full-length continuous performances, it can also support workflows that require natural transitions, preparation movements, idle periods, and longer temporal motion sequences rather than isolated action clips.

Dataset Access

Access to the Apple Arts Studios Full-Length Motion Capture Dataset is provided through the Hugging Face gated dataset access system.

Users must submit an access request before downloading or using the dataset.

All access requests are manually reviewed by Apple Arts Studios.

Approval of a request does not override the terms of the Apple Arts Studios Dataset License. Users are responsible for reviewing and complying with the applicable license conditions before using the dataset.

Access may be granted based on the intended research, development, academic, robotics, AI, animation, or other proposed use of the dataset.

Dataset Limitations

This repository contains full-length, non-segmented motion capture sequences.

Users should be aware of the following characteristics:

Individual files may contain T-pose calibration, idle periods, preparation movements, transitions, repositioning, repeated actions, and settle time.
The motion data is not divided into short action-specific clips.
Users requiring isolated actions may need to perform their own temporal segmentation or trimming.
Motion duration and sequence composition may vary between files.
Unitree G1 motion data represents a robot-oriented retargeted representation of the captured human motion and may require additional processing depending on the user's simulation, control, or robotics pipeline.
Compatibility and retargeting results may vary depending on the target skeleton, robot configuration, software, and processing workflow.

The dataset is intentionally provided in this full-length form to preserve the continuous structure and transitions of the original captured performances.

License and Usage Conditions

This dataset is distributed under the Apple Arts Studios Dataset License.

Access to the dataset does not transfer ownership of the data or grant unrestricted usage rights.

Users must comply with the terms defined in the applicable dataset license.

Commercial Use

Commercial use requires prior written approval from Apple Arts Studios.

Users intending to use the dataset for commercial products, commercial research, production systems, paid services, redistribution, or other commercial activities must obtain the required authorization before use.

Attribution and Redistribution

Users should review the full license terms for requirements relating to:

Attribution
Redistribution
Modification or derived datasets
Commercial usage
Research usage
Dataset sharing
Publication or distribution of derived results

The full Apple Arts Studios Dataset License should be reviewed before downloading or using any portion of this dataset.

Citation

If you use the Apple Arts Studios Full-Length Motion Capture Dataset in research, publications, demonstrations, or other permitted work, please cite the dataset and credit Apple Arts Studios as the dataset creator.

Suggested Citation

Apple Arts Studios. (2026). Apple Arts Studios Full-Length Motion Capture Dataset. Hugging Face.

BibTeX

@misc{appleartsstudios_full_length_motion_2026,
  author       = {Apple Arts Studios},
  title        = {Apple Arts Studios Full-Length Motion Capture Dataset},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/Appleartsstudios/Full_Length_Motion_Capture_Dataset}}
}

For publications using substantial portions of the dataset, users are encouraged to clearly identify Apple Arts Studios as the source of the motion capture data.

Contact and Support

For dataset access, licensing, commercial usage, technical questions, or collaboration enquiries, contact:

Apple Arts Studios Email: hello@appleartsstudios.com

For access requests submitted through Hugging Face, please use the Request Access option available on this dataset repository.

Apple Arts Studios may request additional information regarding the intended use of the dataset before approving access.

Dataset Disclaimer and Ownership

This dataset, published by Apple Arts Studios, was conceived, performed, captured, processed, and curated entirely in-house at our own motion-capture facility for internal research and dataset development.

All motion capture data, processed motion representations, metadata, and associated dataset materials remain the intellectual property of Apple Arts Studios, subject to the terms of the Apple Arts Studios Dataset License.

The dataset is provided for approved research, development, academic, robotics, AI, animation, and related uses. Availability through Hugging Face does not constitute a transfer of ownership or unrestricted permission to redistribute, resell, sublicense, or commercially exploit the dataset.

Users are responsible for ensuring that their use of the dataset complies with all applicable license terms and access conditions.

Downloads last month: 11

Total file size:

575 kB