Today, we are excited to announce that Stable Diffusion XL 1.0 (SDXL 1.0) is available for customers through Amazon SageMaker JumpStart. SDXL 1.0 is the latest image generation model from Stability AI, offering native 1024-pixel image generation at various aspect ratios. It is designed for professional use, providing high-resolution photorealistic images. SDXL 1.0 comes with preset art styles suitable for marketing, design, and image generation across industries. You can easily access and utilize these models with SageMaker JumpStart, an ML hub that offers algorithms, models, and ML solutions to help you quickly start with ML. This post will guide you on how to use SDXL 1.0 models via SageMaker JumpStart.
Stable Diffusion XL 1.0 (SDXL 1.0) is the evolution of Stable Diffusion and represents the next level of generative AI for images. It is capable of generating stunning images with complex concepts and various art styles, including photorealism, surpassing the capabilities of current image models. SDXL, like its predecessor, is highly customizable in terms of parameters and can be deployed on Amazon SageMaker instances.
SageMaker JumpStart provides ML practitioners with a wide range of state-of-the-art models for many use cases such as content writing, image generation, code generation, question answering, copywriting, summarization, classification, and information retrieval. ML practitioners can deploy foundation models to dedicated SageMaker instances in a network isolated environment and customize models using SageMaker for model training and deployment. The SDXL model is now discoverable in Amazon SageMaker Studio and available in multiple regions.
In this post, we will demonstrate how to deploy SDXL 1.0 to SageMaker and use it to generate images using both text-to-image and image-to-image prompts. SageMaker Studio, a web-based integrated development environment (IDE) for ML, will be used to build, train, debug, deploy, and monitor ML models. For more details on how to get started with SageMaker Studio, refer to Amazon SageMaker Studio. Once in the SageMaker Studio UI, you can access SageMaker JumpStart and search for Stable Diffusion XL. Choose the SDXL 1.0 model card, which will open an example notebook. Compute costs are responsible for users, with no associated model cost.
SDXL 1.0 comes in two variations: Closed weight and Open weight. Closed weight SDXL 1.0 offers optimized scripts and containers for faster inference time and can be run on smaller instances compared to the open weight version.
Please refer to the Stability AI’s text to image documentation for more information on the parameters used by SDXL 1.0. Examples in this post are based on the sample notebook for Stability Diffusion XL 1.0, available on Stability AI’s GitHub repo.
In the examples that follow, we will explore the capabilities of SDXL 1.0 models, including superior photorealism, enhanced image composition, realistic face generation, and improved visual aesthetics. We will also demonstrate the use of shorter prompts for creating descriptive imagery, as well as legible text in images.
To generate images using SDXL 1.0, you can provide prompts in text format. The model will generate images based on these prompts. Additionally, you can use style presets to guide the image generation process towards a specific art style. Multiple prompts can be used together, assigning each prompt a specific weight to mix concepts. SDXL 1.0 also supports spatially aware generated images and negative prompts.
We hope this post provides a comprehensive understanding of SDXL 1.0 and how to utilize it with SageMaker JumpStart. Happy image generation!
