NVIDIA drops early preview of Isaac Sim 5.0 and Isaac Lab 2.2

Isaac Sim 5.0 and Isaac Lab 2.2 are NVIDIA's physics-based simulation and learning toolkits, built to help robotics teams train, validate, and test machines in virtual worlds before touching physical hardware.

NVIDIA rolled out early developer previews for both releases on GitHub. If you are building robot perception systems, testing manipulation, or scaling reinforcement learning, you can get early access right now.

What you can do with it

The biggest shift is accessibility. Isaac Sim-specific extensions are now open source and hosted in a new public GitHub repository.

Under the hood, Isaac Sim 5.0 adds advanced synthetic data generation pipelines, new ready-to-use robot models, and improved sensor simulation. It also standardizes ROS2 interfaces, making it much easier to plug into standard robotics stacks without messy glue code.

Isaac Lab 2.2 focuses heavily on robot learning and motion. It brings enhanced synthetic motion generation, deeper integration with Omniverse Fabric for fast data exchange, and new simulation features like a tensorized suction cup gripper.

Alongside the preview, NVIDIA highlighted the Isaac GR00T-Dreams blueprint. Designed specifically for humanoid robots, it generates massive amounts of synthetic trajectory data so robots can adapt to new environments with minimal human demonstration data.

How to get started and current limits

Both releases are available right now on GitHub as early developer previews.

You can clone the repositories, inspect the extensions, and join ongoing discussions on the official NVIDIA developer forums or Discord.

Keep in mind that these are still early preview builds. APIs and workflows may change as NVIDIA collects community feedback, so they are best suited for testing pipelines rather than live deployments.

Why it matters

Real-world robot data is notoriously slow and expensive to collect. Teaching a humanoid robot to handle edge cases usually takes thousands of human demonstrations that smaller teams simply cannot afford.

By open-sourcing more of Isaac Sim's core extensions and streamlining synthetic data pipelines on GitHub, NVIDIA is removing major friction for researchers and independent builders. While stability may still be rough around the edges, robotics developers should start experimenting with these workflows today.