Data
The world's largest studio-grade human motion dataset
At Uthana, we treat human motion as structured data — not as an artifact of a capture session, but as a queryable, programmable resource. Our dataset of 75,000+ studio-grade motion clips is the foundation layer for everything we build: generative models that create new motion from text, retargeting systems that adapt to any skeleton, and real-time inference engines that run at millisecond latency.
What's in the dataset
Specifications & composition
Every clip is cleaned, labeled, and physics-annotated in-house. Built for production and research from day one.
Total clips
50,000+ and growing weekly
Frame rate
60–120 FPS per clip
Clip duration
0.5 – 45 s (median 4.2 s)
Skeleton joints
22 – 65 joints, retargetable to any rig
Capture hours
500+ hours of mocap
Export formats
FBX · GLB · BVH
License
Commercial use per plan tier
Sourcing
Performer-consented, fully owned
Why this dataset is different
Not just motion capture — motion infrastructure
Standard mocap datasets give you clips. Uthana gives you a programmable motion layer built for production at scale.
Scalable by design
Traditional datasets are fixed collections — what you download is what you get. Uthana’s foundation models generate unlimited new motion on demand, so the dataset grows with your needs rather than constraining them.
Production-ready, not research-only
Every clip ships with proprietary IK retargeting that maps to any skeleton, export in FBX/GLB/BVH, and real-time inference under 15ms. This isn’t data you clean for six months before using — it’s ready for your pipeline today.
Controlled via natural language
Describe what you need in plain text and get motion back. Query the dataset semantically rather than browsing folders of numbered files. Combine, stitch, and loop clips programmatically through our API.
Diverse beyond any one application
The same motion data that trains a humanoid locomotion controller in MuJoCo also benchmarks gait asymmetry in a sports science lab and populates a crowd scene in Unreal. One dataset, every domain.
Access the motion data
From single API calls to bulk dataset licensing — choose the access pattern that fits your workflow.
API access
Generate on demand
Query our foundation model via API. Describe what you need in natural language and get production-ready motion in milliseconds. GraphQL + Python SDK.
Custom datasets
Tailored to your needs
Download curated subsets of relevant data in FBX, GLB, or BVH. Includes all metadata, annotations, and physics labels. Commercial license available.
Enterprise
Build models with us
Bring your own data, or create more with us. Private models, data siloing, on-prem deployment, and dedicated engineering support are available.
Studio & partner toolchain
Join us in co-developing the future of AI-driven animation.
Style transfers
Apply your game style to generic motions across your team's models.
Integrations
Leverage the Uthana platform in your development pipeline or game environment.
Data siloing
Train bespoke models, or organize and label your data with our ML tools.