Skip to content

Dataset · tencent

Large-scale style dataset for image generation

tencent/MegaStyle-1.4M

1.4M English text-image pairs with high artistic style consistency.

What's inside

The dataset from Tencent contains 1 to 10 million examples (1.4 million pairs in the base version) including images and English text descriptions. The images were generated using Qwen-Image by combining 170k style prompts and 400k content prompts to maintain a consistent style.

What you can do with it

  • Training text-to-image models for style transfer and precise style adherence
  • Training style LoRA adapters for image generators
  • Researching style and content disentanglement in multimodal models
Tasks
Text → image
Languages
English
Size
1M–10M rows
Downloads this month
1.6K
Likes
55

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("tencent/MegaStyle-1.4M")