Uthana is an AI platform for human motion, with its core positioning being "Turn observation and intent into 3D motion," that is, providing models and data for physical AI (robots, simulation, 3D characters).
In terms of use, it can transform observations and intent into structured 3D human motion, used to generate animations, train models, and drive responsive virtual human agents.
Main capabilities include:
- 2D video to 3D motion: Convert 2D video into editable 3D motion data that can be used directly for animation or training.
- Text to 3D motion: Generate editable 3D character motion by describing actions, sequences, rhythm, poses, and style.
- Real-time/responsive motion generation: Generate responsive motion based on direction, speed, and trajectory.
- Motion retargeting: Transfer human motion across different skeletons, characters, and robot morphologies without manual reconstruction; it can also automatically complete rigging preparation for unrigged 3D characters and convert human motion into robot-specific joint representations.
- Motion data for training: Provide licensing of high-quality captured motion data, or transform real-world video into structured 3D motion data, annotations, and model-ready datasets.
- Dataset capabilities: Includes the "Studio Motion Dataset" collected by an AAA production studio using a 120-camera motion capture system; supports extracting motion data from video and can add action labels, time segments, joint trajectories, contact states, and custom metadata to match training specifications.
- API and integration: Complete motion generation, video motion extraction, cross-body retargeting, and automatic rigging through a unified GraphQL API; provide animation formats for production tools and tabular formats for training pipelines, with specific fields, skeleton conventions, and units explained in the dataset documentation before delivery. It can be integrated into engines, pipelines, or platforms through the GraphQL API and public plugins.
In terms of usage, users can start generating motion through the platform, or embed the capabilities into their own products through the GraphQL API and public plugins. The documentation states that it provides a unified API from prototype to production and supports collaboration with simulation teams to develop interactive human agents.
Applicable scenarios include: supporting robot training, interactive simulation, world models, and 3D experiences; generating, retargeting, and delivering production-grade character animation for games and physical AI; supplementing multimodal training data with structured 3D motion and action labels; and enabling AI-native game platforms to embed motion generation, rigging, and retargeting capabilities; it also supports direct collaboration on custom datasets, morphology systems, and simulation projects.
In terms of limitations, the official text does not provide specific pricing, free trial policy, supported platforms, or system requirements; this information is not currently available. The exact fields, skeleton conventions, and units of the dataset delivery format need to be confirmed according to the specific dataset documentation.
