Skip to content

Dataset · showlab

Paired images across 22 art styles

showlab/OmniConsistency

Dataset of source-to-stylized image pairs with text prompts for training style transfer.

What's inside

Contains aligned image pairs: an original (photo or line sketch) and its version in one of 22 art styles, supplemented with a text style description. Dataset size is from 1,000 to 10,000 records. File format and prompt language are not specified in the card.

What you can do with it

  • Training style transfer and image-to-image generation models
  • Training LoRA for specific art styles with text control
Size
1K–10K rows
Downloads this month
10.9K
Likes
39

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("showlab/OmniConsistency")