Stability AI is a tech startup developing the “Stable Diffusion” AI model, which is a complex algorithm trained on images from the internet. Following a test version available to researchers, the company has officially released the Stable Diffusion model, which can be used to create images from text prompts. Unlike Midjourney and other models/generators, Stable Diffusion aims to create photorealistic images first and foremost — something that has already led to controversy over “deepfake” content. However, it can also be configured to mimic the style of a given artist.
Stable Diffusion is unique because it can run with a typical graphics card, instead of using remote (and expensive) servers to generate images. Stability AI recommends using NVIDIA graphics cards right now, but full support for AMD and Apple Silicon is in the works.
Stable Diffusion has a ‘Safety Classifier’ mode that attempts to block offensive images from being generated, but because the model is open-source, it can be turned off when running on a PC. Web-based generators often prevent people from using prompts that mention certain words or phrases, to prevent creating images that could be used to deceive or harm others. For better or worse, Stable Diffusion can create more types of images than is possible with most other services.
The Stable Diffusion model is designed to be used with an accompanying generator, which is the actual interface used for typing prompts and changing other options. The Stable Diffusion Dream Script is one generator that can run locally on a computer, either through a command-line interface or a local web server. There’s also an in-development plugin for generating images from inside Photoshop, and web-based versions are already popping up. Check out our guide to running Stable Diffusion locally for the step-by-step process.
Given the open-source licensing model of Stable Diffusion, and its impressive generation abilities, it’s likely that most AI generators will adopt the new model.
Source: Stability AI