This video introduces the key new features released in Rerun 0.14, a visualization tool for multimodal data. The presenter, Paul, walks through four major updates using a live robotics teleoperation demo. These updates include the highly-requested Light Mode for the UI, a built-in URDF loader that simplifies robotics visualization, and live encoded video streaming to drastically reduce data logging sizes. The most significant update is an underlying architectural change called "Tagged Components," which enhances the flexibility of Rerun's data model by allowing multiple components of the same type to be logged to a single entity without conflicts.
What's New in Rerun 0.14: Key Features and Improvements
00:00:41 New Light Mode: Rerun now includes a light mode interface, a long-requested alternative to the default dark mode. It can automatically respond to system settings or be set manually.
00:00:58 Integrated URDF Loader: Previously requiring an external tool, the data loader for URDF (Unified Robot Description Format) files is now built directly into Rerun. This allows for seamless, automatic logging and live visualization of robot models during operation.
00:01:35 Live Encoded Video Streaming: Users can now log live encoded video streams instead of individual raw images. This significantly reduces file sizes (from gigabytes to megabytes) for long recordings, which is crucial for robotics data collection and training, while maintaining smooth playback.
00:02:44 Flexible Data Recording: The new multi-sink API allows developers to simultaneously stream data to a live viewer and save it directly to a file. This automates the data capture process, which is ideal for creating training datasets without manual intervention.
00:03:24 Tagged Components for Advanced Data Modeling: A major architectural improvement where components are tagged with their field name and archetype. This prevents data conflicts and allows users to log multiple components of the same type (e.g., multiple Position3D components) to a single entity, making the data model more powerful and flexible. This change also groups components by archetype in the UI and removes the need for "indicator" components, reducing clutter and file size.