LAION Releases 80 Million-Video Open Dataset for Multimodal AI Research

LAION has published LAION-BVD, an open video dataset of 80 million downloaded videos totaling 10 million hours, indexed via 1.3 billion platform-specific URLs. The non-profit released the dataset on August 27, 2026, through its GitHub repository and project page. Licensed for research use only — not commercial — BVD supports multimodal pre-training across video, audio, and image modalities, addressing a persistent scarcity of open video corpora at this scale.