Stability AI introduced SDXL 0.9: An updated neural network for creating photorealistic images
Stability AI has published an updated version of its neural network for generating images from a text description, called SDXL 0.9. The previous version of the model has been removed from the blog, but the new version promises to create better and more detailed photorealistic images.
Breakthrough in the creative use of generative image AI
SDXL 0.9 has been hailed by developers as a breakthrough in the creative use of generative artificial intelligence for image creation. Stability AI backed up its words with paired examples of images created by the new model and the previous version of Stable Diffusion XL: photorealistic images of a wolf, an alien in Las Vegas, and a human hand with a coffee cup. In each case, SDXL 0.9 provides more detailed and believable pictures.
Increasing parameters for success
Stability AI noted that successful results were achieved due to a significant increase in the number of parameters in the updated model. It is important to note that SDXL 0.9, like previous platforms, can be run on a home computer. This model requires 16 GB of RAM and an NVIDIA GeForce RTX 20 or higher graphics card with 8 GB of memory. It is compatible with Windows 10/11 and Linux operating systems.
Availability and future plans
The updated SDXL 0.9 model will soon be available on the Stability AI Clipdrop web service and the DreamStudio app, notes NIX Solutions. An open source release of SDXL 1.0 is also scheduled for mid-July.
As a result, Stability AI’s SDXL 0.9 promises an impressive breakthrough in photorealistic imaging, offering higher quality and more detail. This update opens up new perspectives for the creative application of generative artificial intelligence and is an important step in the development of this field.