Skip to content

Dataset · DropletX

10M video generation descriptions and metadata

DropletX/DropletVideo-10M

Large-scale open dataset for text-to-video and image-to-video tasks focusing on consistency.

What's inside

The largest open corpus for video generation with spatiotemporal consistency, containing between 10 and 100 million records. The repository features English text descriptions and tabular metadata, along with a filtered premium subset, DropletVideo-1M, with high aesthetic and image quality ratings.

What you can do with it

  • Training and fine-tuning text-to-video and image-to-video models
  • Improving spatiotemporal consistency in video generation
  • Filtering and selecting quality video pairs for training
Tasks
Image → video, Text → video
Languages
English
Size
10M–100M rows
Downloads this month
194
Likes
34
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("DropletX/DropletVideo-10M")