qizheng-ai2026openmdw-license-agreement-version-1.1

BridgeData2 Subset Synthetic Captions

A subset of BridgeData V2 containing robot manipulation video clips paired with synthetic natural-language captions, designed for supervised fine-tuning and evaluation of text-to-video, image-to-video, and video-to-video generation models.

Downloads0
Episodes1273

Why This Matters for Physical AI

This dataset enables training and evaluation of vision-language-action models and world-generation models for robot manipulation through paired video-caption data, supporting scalable robot learning and video generation research.

Technical Profile

Modalities
rgb
Robot Embodiments
WidowX 250
Environment
labkitchen
Task Types
manipulationrobot-manipulation
Episodes
1273
Data Format
MP4 + TXT + JSONL
Annotation Types
language_instructionssynthetic_captions
License
openmdw-license-agreement-version-1.1
Part of the BridgeData V2 family

Access

Need custom rgb data?

Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets