Dataset · LanguageBind
Pexels videos and annotations for Open-Sora Plan
LanguageBind/Open-Sora-Plan-v1.1.0
Around 100k–1M 1080p videos and text descriptions for training video generation models.
What's inside
Contains stock video clips from Pexels compressed to 1080p resolution, along with text annotations. Due to platform limits, archives are split into five parts per folder and must be concatenated before extraction. The authors note that filenames may not match the original annotation file after resolution changes.
What you can do with it
- Training and fine-tuning text-to-video generation models
- Training video encoders and automated video captioning systems
- Preparing high-resolution 1080p visual data
- Size
- 100K–1M rows
- Downloads this month
- 174.5K
- Likes
- 56
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("LanguageBind/Open-Sora-Plan-v1.1.0")