Bones Studio provides enterprise‑grade, IP‑clean multimodal motion capture datasets that synchronize full‑body, finger, facial, voice, and 3D scene data frame‑by‑frame. The datasets include semantic labeling, multi‑camera video, and environment reconstructions, and can be licensed or custom‑engineered for robotics, AI, and game development applications.
Funding
Funding not disclosed
Founders
Product
Problem
Developing realistic humanoid control, imitation learning models, and lifelike character animation requires large volumes of accurately captured human motion data across multiple modalities. Existing datasets often lack synchronization between body, hand, facial, voice, and environmental information, leading to gaps in training data and limiting model performance.
Solution
Bones Studio offers enterprise‑grade, IP‑clean multimodal datasets that capture full‑body motion, finger articulation, facial expressions, voice, and 3D scene reconstruction in perfectly synchronized frames. Each dataset includes semantic labeling, multi‑camera reference video, and scanned environment models, providing a coherent ground‑truth representation of human behavior. The data is ready for licensing or can be custom‑tailored to fit specific pipeline requirements, including format, quality, and scalability. By delivering synchronized, scene‑aware recordings, Bones enables researchers and developers to train humanoid control policies, world models, and high‑fidelity game characters with reduced sim‑to‑real gaps. The platform also supports collaborations for bespoke dataset creation to address niche application needs.
Target Audience
Primary customers are robotics research labs, AI teams building humanoid control or world‑model systems, and game development studios seeking high‑fidelity motion and animation data.
Features
- Full‑body motion capture combined with precise finger articulation and facial capture in a single synchronized recording
- Integrated voice recordings and behavioral context aligned frame‑by‑frame with motion data
- 3D scene reconstruction and object tracking, providing environment‑aware motion datasets
- Semantic labeling and multi‑camera reference videos for comprehensive ground‑truth validation
- IP‑clean licensing model allowing unrestricted commercial use
- Custom dataset engineering services to match specific data structures, formats, and quality standards