BridgeData2 Subset Synthetic Captions: Robot Manipulation Video Clips

A subset of BridgeData V2 packaged with short robot-manipulation video clips and synthetic video captions. It was created by NVIDIA and last updated on June 1, 2026. The dataset is intended for workflows involving text-to-video, image-to-video, and video-to-video generation of robot manipulation scenes.

Use Cases

• Supervised fine-tuning of video generation models based on robot-manipulation video clips.
• Prompt generation for text-to-video models based on synthetic video captions.
• Evaluation of image-to-video generation models on robot manipulation scenes.
• Video-to-video generation workflows using paired video clips and synthetic captions.

Strengths

• Dataset includes synthetic video captions, providing text-video pairs.
• Derived from BridgeData V2, a large-scale robotics dataset.
• Specifically packaged for text-to-video, image-to-video, and video-to-video generation workflows.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Derived from BridgeData V2.

Freshness

Last updated 2026-06-01 05:40:45; freshness should be verified.