What is Stable Diffusion?

What is Stable Diffusion?
Rory Greig Updated on by

Video Gamer is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Prices subject to change. Learn more

Continuing our deep-dive into image generating AIs we’re answering the question of what is Stable Diffusion. Like other popular visual AIs such as Midjounrey or Dall-E, Stable Diffusion is a text-to-image model. This means it will create an image in response to a prompt given by the user.

It gets its name from the ‘diffusion model’ of deep learning algorithms. This is essentially a way of training generative AIs. Stable Diffusion is fed a tonne of data, in this case, images, and uses an amalgamation of this data to generate new unique images. The maths used to calculate and compute this looks like diffusion in physics, hence the name.

The AI was developed by Stability AI and can be accessed here.

EXCLUSIVE DEAL 10,000 free bonus credits

Jasper AI

On-brand AI content wherever you create. 100,000+ customers creating real content with Jasper. One AI tool, all the best models.

Copy AI

Experience the full power of an AI content generator that delivers premium results in seconds. 8 million users enjoy writing blogs 10x faster, effortlessly creating higher converting social media posts or writing more engaging emails. Sign up for a free trial.
ONLY $0.01 PER 100 WORDS

Originality AI detector

Originality.AI Is The Most Accurate AI Detection.Across a testing data set of 1200 data samples it achieved an accuracy of 96% while its closest competitor achieved only 35%. Useful Chrome extension. Detects across emails, Google Docs, and websites.

What was the Stable Diffusion model trained on?

Stable diffusion was trained on a publicly available dataset called LAION-5B. A German non-profit called LAION developed the resource which is built from 5 billion image-text pairs, scraped for the internet. It seems that with this model certain domains make up a disproportionately high percentage of the origins of the training images.

This means that Stable Diffusion was trained on many different images but a significant amount of its data came from places such as Pinterest, WordPress and Flickr.

Each image and text pair taken from LAION-5B is filtered and classified so the AI can know what defines each one. This is part of how it can then generate new images based on language.

With the recent Midjourney job action restricted message seemingly removing free credits from Midjourney entirely, people are looking for AI image-generating alternatives and Stable Diffusion is a great one to give a go. For now, Firefly is free too, but you have to be accepted onto the Beta.

Frequently asked questions

Can artists opt out of having Stable Diffusion trained on their art?

In their FAQs Stable Diffusion acknowledges that artists could not opt-in or opt out of having their images be part of the training data.

Can anyone use Stable Diffusion?

Yes – you can use Stable Diffusion for free for both commercial and non-commercial purposes.