The update brings improvements to SD's text-to-image diffusion models, includes a powerful image upscaler, updates its Inpainting model, and more.
Stability AI has announced the release of Stable Diffusion 2.0, a new and improved version of its multi-functional open-source AI Stable Diffusion. The release delivers a large number of useful enhancements to SD, improving its text-to-image diffusion models, bringing a powerful new image upscaler, updating its Inpainting model, and more. According to the developers, the upgraded version was designed to "serve as the foundation of countless applications and enable an explosion of new creative potential".
And here's the list of improvements and new features Stable Diffusion 2.0 brings:
- New Text-to-Image Diffusion Models: Stable Diffusion 2.0 includes robust text-to-image models trained using a brand-new text encoder, which greatly improves the quality of the generated images compared to earlier V1 releases, enabling one to generate images in 512x512 and 768x768 resolutions.
- Super-resolution Upscaler Diffusion Models: The release brings an Upscaler Diffusion model that enhances the resolution of images by a factor of 4.
- Depth-to-Image Diffusion Model: A new diffusion model capable of inferring the depth of an input image and then generating new images using both the text and depth information.
- Updated Inpainting Diffusion Model: A new text-guided inpainting model, fine-tuned on the new Stable Diffusion 2.0 base text-to-image, which makes it easy to switch out parts of an image.