Dataset · nyanko-devs
Over 10M anime illustrations with Danbooru tags
nyanko-devs/danbooru2026
A massive anime art corpus with detailed user tags for image generation and classification.
What's inside
The repository gathers 10 million to 100 million anime illustrations from the Danbooru imageboard, annotated by the community in English and Japanese. The project is actively in progress (WIP) and features an extensive tagging system for characters, styles, and scene details.
What you can do with it
- Training from scratch and fine-tuning generative models (text-to-image and image-to-image) for anime style
- Training multimodal models and taggers for illustration classification
- Tasks
- Image classification, Image → image, Text → image
- Languages
- English, Japanese
- Size
- 10M–100M rows
- Downloads this month
- 26.5K
- Likes
- 51
How to load it
from datasets import load_dataset
ds = load_dataset("nyanko-devs/danbooru2026")