This latest iteration marks a revolutionary advancement in the world of text-to-image models.
Open generative AI pioneer Stability AI has launched its newest model, Stable Diffusion XL (SDXL) 1.0. This advanced text-to-image model is now featured on Amazon Bedrock, an Amazon Web Services (AWS) service that offers foundational models from top AI startups and Amazon Titan via an application programming interface (API).
The allure of AI art generators has surged in recent times. These innovative tools harness the power of machine learning (ML) algorithms and vast datasets to create mesmerizing, often dreamlike images. From artists and designers to casual enthusiasts, many are captivated by the distinctive and unexpected results AI-generated art offers. These generators, whether text-to-image or image-based, draw inspiration from existing artworks, visual styles and imagery.
The unveiling took place at the AWS Summit New York, highlighting Amazon’s dedication to offering powerful text-to-image models like SDXL 1.0. Emad Mostaque, CEO of Stability AI, expressed the company’s commitment to collaborating with AWS to deliver top-notch solutions for developers and clients.
Developers can now access SDXL 1.0 via Stability AI’s API, its GitHub page, and consumer apps like Clipdrop and DreamStudio. Additionally, the model is available on Amazon SageMaker JumpStart, a hub for ML solutions.
Enhanced image quality and generation speed
Building on its research-only predecessor, SDXL 0.9, which debuted in June, the full version of SDXL has since undergone significant improvements, solidifying its position as the leading open image generation model on a global scale. The model’s latest iteration can produce images with richer colors, sharper contrast and enhanced lighting and shadows. The revamped imaging process also ensures rapid outputs, churning out full 1-megapixel (1024×1024) resolution images in mere seconds across diverse aspect ratios.
A unique feature in the pipeline allows users to fine-tune the model using just five images, enabling specialized image generation on specific people, products and more. This feature is set for release soon.
Powerful parameter count
SDXL 1.0 stands out with its extensive parameter count, featuring a 3.5 billion parameter base model and a 6.6 billion parameter model ensemble pipeline. Since its beta launch on April 13, the AI art platform, ClipDrop, has seen a remarkable level of user engagement, creating an impressive 35 million images using the SDXL model. Stability AI’s Discord community mirrors this fervor, averaging 20,000 images daily via SDXL 1.0.
Swami Sivasubramanian, Vice President of Database, Analytics and Machine Learning at AWS, emphasized the importance of model choice in generative AI. He praised the addition of SDXL 1.0 to Amazon Bedrock, noting it provides customers with a top-tier text-to-image model for innovative AI applications.
In 2022, Stability AI chose AWS as its primary cloud provider and was among the first to feature its models on Amazon Bedrock. The release of SDXL 1.0 strengthens the bond between the two tech giants, aiming to simplify and accelerate the development of generative AI applications.
This launch is one of many recent achievements for Stability AI, which also introduced a new developer platform and released Stable Doodle, a sketch-to-image tool that produced over 3 million images within a week of its debut.
Also read:
- Discovering the Best AI Art Generators 2023
- How Picsart Uses AI to Empower Creators
- Drawing without a Paintbrush: An Evolution of Artificial Intelligence
Header image courtesy of Pexels
Press release link: https://stability.ai/press-articles/stable-diffusion-xl-1-featured-amazon-aws-bedrock
