Stability AI’s new Text to Image Model
An impressive upgrade
When Stability AI announced the new Stable Cascade model, it was not billed as an upgrade to Stable Diffusion XL, but as a new method in itself.
Without going into the technical detail, the new model uses a radically different architecture to Stable Diffusion 1.5 and SDXL. But it retains the ease of use associated with SDXL. In fact, in testing, the best results seem to result from very simple prompts.
The video below discusses the earliest tests and the differences between SDXL and Stable Cascade.
But what is not immediately clear is the stunning quality of the images that Stable Cascade can produce.
Super-powered Generative AI
With the correct settings, amazing results can be obtained. Below is just one example achieved with a simple prompt:
renaissance theme, feminine beauty, female protagonist, vibrant color
Results like this do need higher than average steps. Whilst SDXL operates best at around 1 megapixel (1024 x 1024 px for example), Stable Cascade can readily scale to 2 or even 4 megapixels.
Twitter displays this image in it’s original resolution – 2048 x 2048 px.
Renaissance beauty – Stable Cascade with no upscale #stablecascade pic.twitter.com/q85PLYXXFS
— Pixovert (@pixovert) February 15, 2024