metadata

task_categories:
  - text-to-video
  - text-to-image
  - video-classification
  - image-classification
language:
  - en
size_categories:
  - 1K<n<10K
license: agpl-3.0

TempoFunk Small

7.8k samples of metadata and encoded latents & prompts of random popular preview videos from shutterstock.

Data format

Video frame latents
- Numpy arrays
- 120 frames, 512x512 source size
- Encoded shape (120, 4, 64, 64)
CLIP (openai) encoded prompts
- Video description (as seen in metadata)
- Encoded shape (77,768)
Shutterstock video metadata as JSON