Simulation-ready models for
MuJoCo
Convert 3D models, images, or text into models with accurate physics properties for MuJoCo. AI estimates mass, friction, contact parameters, and generates collision geometry automatically.
The problem
Why existing workflows fall short.
MuJoCo XML authoring is manual
Defining geoms, bodies, and joints with accurate physical parameters in MJCF XML requires domain expertise and per-object tuning.
Physics estimation guesswork
Estimating mass, friction, and contact parameters for arbitrary objects is time-consuming. Inaccurate values lead to unrealistic sim behavior.
Scaling object diversity
Training generalizable manipulation policies needs thousands of varied objects. Creating each one manually doesn't scale.
How Rigyd helps
AI-native infrastructure that automates the hard parts.
AI-powered contact parameters
Rigyd estimates friction, restitution, and condim values using material identification, calibrated for MuJoCo's contact model.
Accurate mass and inertia
Volume estimation combined with material density lookup produces mass and inertia values within domain randomization variance ranges.
Bulk object generation
Convert entire object datasets at once. Enterprise API supports high-throughput pipelines for building diverse training object sets.
faster than manual physics annotation
mass accuracy vs measured (within DR ranges)
saved per 1,000-object project
Build better MuJoCo training environments
Drop in a 3D model, image, or text description and get physically accurate assets for MuJoCo in minutes.
Frequently asked questions
Does Rigyd output MJCF XML directly?
Rigyd emits OpenUSD plus MJCF XML for MuJoCo workflows. Geometries, bodies, contact parameters (solimp, solref, condim), and inertia tensors are written to match MuJoCo's physics model exactly.
How are MuJoCo contact parameters estimated?
Rigyd identifies per-region materials using multi-view analysis and maps them to calibrated friction, restitution, and condim values, tuned for MuJoCo's contact solver rather than generic defaults.
Is Rigyd suitable for MJX and Brax training?
Yes. MJX consumes the same MJCF, so Rigyd-generated assets run in MuJoCo, MJX, and Brax pipelines without modification. Bulk conversion makes large, diverse object sets practical for RL training.
Does Rigyd export MJCF XML directly?
OpenUSD is the canonical output; MJCF is generated on-demand from the same source via Rigyd's converter. Conversion preserves mass, inertia tensors, friction coefficients, and condim values exactly. Each asset becomes a standalone MJCF model file, or composes into a shared MJCF worldbody for multi-object scenes, both formats validate cleanly with mujoco compile.
How are MuJoCo contact parameters (condim, friction, solref, solimp) populated?
Friction (μ, μ2, μ3 for tangential and torsional) is set per identified material. Condim defaults to 3 for typical rigid contact and 4 for objects requiring torsional friction (cylindrical graspables, screws). Solref and solimp use MuJoCo's recommended defaults for rigid-body contact; override per-asset if you need softer contact for compliant grippers. For research workflows that need tighter contact-parameter control, Rigyd exposes condim, solimp, and solref as per-asset overrides via the dashboard, useful for sysID experiments where the simulator needs to match a specific real-robot contact profile. Overrides persist across re-runs of the same asset and version-control cleanly in git as MJCF XML.
Related reading
In-depth guides on robotics simulation, OpenUSD, and SimReady assets.
9 JUL 2026
Scaling Simulation Asset Libraries Beyond Curated Inventory
Curated SimReady libraries are a great starting point but a hard ceiling. Here is why fixed inventories limit robotics simulation at scale, and how on-demand asset generation closes the gap.
2 JUL 2026
Domain Randomization for Robotics Training: Asset Diversity at Scale
Domain randomization makes trained policies transfer to the real world, but it only works if your asset library is diverse enough. Here is how asset diversity drives randomization, and how to generate it at the scale training needs.
30 JUN 2026
Building a Scalable Embodied AI Asset Pipeline: From Raw Data to Simulation
A practical look at the stages of an embodied AI asset pipeline, from raw 3D data and reference inputs to physics-ready simulation assets, and what changes when you need to produce thousands of them instead of a handful.
Explore more
SimReady for Isaac Sim
OpenUSD assets validated for NVIDIA Isaac Sim
Models for MuJoCo
Physically accurate assets for robot learning
Assets for Gazebo
Simulation-ready models for ROS 2
Warehouse Simulation
50,000+ SKUs need physics properties
Humanoid Robots
Whole-body interaction training assets
Sim-to-Real Transfer
Close the gap with accurate physics data