LAION
The non-profit behind LAION-5B, the open image-text dataset a great deal of open generative AI was trained on.
Highlights
- Extensive Datasets: Offers large datasets like LAION-400M and LAION-5B.
- Multilingual Support: Includes multilingual CLIP-filtered image-text pairs.
- AI Models: Provides access to advanced AI models like Clip H/14.
- Aesthetic Filtering: Features subsets filtered for aesthetically pleasing images.
- Resource Efficiency: Encourages the reuse of existing datasets and models.
External link — opens laion.ai in a new tab. LAION is a third-party product; we are not affiliated with it.
About LAION
What it is
LAION is a non-profit that publishes open machine learning resources — most significantly LAION-5B, a dataset of billions of image-text pairs scraped from the web, which was the training data behind Stable Diffusion and much of the open image-generation ecosystem. It also releases models and tooling.
Why it's different
Its importance is hard to overstate and rarely acknowledged: open generative AI exists in the form it does largely because this dataset was published rather than kept. That is the argument for it. The honest account has to include the rest. LAION-5B is web-scraped, which means it contains copyrighted work whose creators did not agree to it, and in 2023 researchers found child sexual abuse material in it, after which it was withdrawn and re-released in a filtered form. Anyone building on open image models is building on this lineage, and the licensing questions currently in court are about exactly that.
How people use it
It matters to researchers and to anyone training or fine-tuning image models, where the datasets are the starting point. For everyone else it is essential context rather than a tool: when a commercial vendor like Bria sells "license-clean training data" at a premium, this is the thing they are selling an alternative to, and understanding that makes the price legible.
Written by the n3os team. We are not affiliated with LAION.
This listing was written from public information, without LAION’s involvement. If you own it and something here is wrong — or you would rather not be listed at all — email us and we will correct or remove it.
Get the ones worth knowing about
We write one of these for every tool worth the trouble. Get the new ones, plus what we have found genuinely useful lately.
Your address goes to Buttondown, who send the emails on our behalf. One click unsubscribes, and the list is never sold or shared.