LAION-BVD: A 10-Million-Hour Open Video Dataset for Multimodal Pre-training

Andreas Hochlehnert · hf · 2026-08-26

LAION-BVD is a large-scale open video dataset with 10 million hours, enabling multimodal pre-training across video, audio, and image modalities. It features synthetic captions and demonstrates strong benchmark performance.

Original post →

More from Multimodal

• Pavo launches AgnesVideo 2.5 with free tier to cut AI short drama costs — 新智元 · 2026-08-26
• Open Source Discord Plugin: Interrogate Images & Generate Prompts via Krea2 — Ok_Clothes9170 · 2026-08-26
• MiniMax H3 Lipsync Workflow: 1.4MP Takes 3-4 Hours on 5090 — TheDerminator1337 · 2026-08-26
• Redditor shares AI-generated anime series 'Realmz of the Redeemers' ep 2 — Kindly_Poet_7878 · 2026-08-26
• Captain America x Harry Potter AI video on a single RTX 3070: 10s clip in ~10 minutes — luka06111 · 2026-08-26
• Redditor shares dark fantasy AI-generated video clip — Illustrious_Fee_3676 · 2026-08-26

Multimodal channel →