Skip to content

Dataset · artplus

Multi-layer transparent images with style descriptions

artplus/PrismLayersPro

20,000 curated multi-layer images with transparency and text style descriptions.

What's inside

PrismLayersPro contains 20,000 high-quality multi-layer images with transparency (metadata indicates a range of 10,000 to 100,000) and rewritten style descriptions. It is a human-curated subset of the larger 200,000-example PrismLayers dataset generated by the MultiLayerFLUX model. The data is divided into 21 splits by visual style.

What you can do with it

  • Training generative models to create layered images with transparency
  • Training LoRAs for specific visual styles across 21 categories
  • Developing layer editing and compositing tools
Size
10K–100K rows
Downloads this month
1.9K
Likes
32

How to load it

from datasets import load_dataset

ds = load_dataset("artplus/PrismLayersPro")