---
license: openrail
---

A sample from the [Laion 6.5+ ](https://laion.ai/blog/laion-aesthetics/) image + text dataset. You can see
some samples [here](http://captions.christoph-schuhmann.de/2B-en-6.5.html). 

The samples are resized + center-cropped to 64x64x3 and the .npz file also contains CLIP embeddings. 

TODO: add img2dataset script.

The data can be used to train a basic text-to-image model.