Stability AI has released a new image generation model, Stable Diffusion XL 1.0

Stability AI has released the Stable Diffusion XL 1.0 text-to-image conversion model, which it calls its "most advanced" version to date.

Here's What We Know

Stability AI's head of applied machine learning, Joe Penna, revealed that Stable Diffusion XL 1.0 contains 3.5 billion parameters. The model can produce full resolution 1 MP images "in seconds" with multiple aspect ratios.

According to the creators, the algorithm delivers "more vibrant" and "accurate" colours, as well as better contrast, shadows and lighting than its predecessor.

Stable Diffusion XL 1.0 supports shading, recolouring and "image-to-image" cues. This means users can give the model an image as input and edit it.

According to Stability AI, the model understands complex, multi-part instructions given in short prompts. Previous models required longer text prompts.

The developers also taught Stable Diffusion XL 1.0 how to generate clear and readable text.

Stability AI acknowledged that the model could generate toxic and biased content. However, the developers have endeavoured to tweak it to reduce the likelihood of generating such images.

Stable Diffusion XL 1.0 is distributed open source on GitHub in addition to Stability's API and consumer apps like ClipDrop and DreamStudio.

Source: TechCrunch