Voxel51 3.0 Platform Enhances Physical AI Data Management

The platform specifically targets teams that are training models using sensor data from robots, vehicles, and factories. The new release introduces Voxel51 Agent, a data provenance layer, and offers free self-service access.

Voxel51 3.0 expands beyond its previous focus on visual data with three core additions. It now supports time-series multimodal data, allowing teams to view multiple sensor streams on a single timeline and use segment-level embeddings to find specific moments within long recordings.

The platform also features unified annotation for 2D, 3D, video, and multimodal data, including automated quality assurance, annotator metrics, and human-in-the-loop review. Agentic Labeling leverages vision-language models to generate labels from natural-language instructions and examples.

Supported data modalities include MCAP, images, video, 3D point clouds, meshes, DICOM, NIfTI, geospatial data, and audio. Voxel51 Search functions as a persistent hybrid index that automatically indexes datasets and computes embeddings, enabling search from the outset.

Semantic retrieval can find visually or semantically similar moments, even without shared labels, while lexical and structured queries handle exact terms, timestamps, sensor fields, and metadata. Vector retrieval is powered by LanceDB.

The free tier of Voxel51 3.0 includes Voxel51 Agent, 30,000 inference requests, and 30 compute hours per month. However, the free tier is capped at three datasets of 10,000 samples each and 20 GB of storage.