Click on the blue "Flying Snow Ruthless" above to follow my official account, set a star, and read the article as soon as possible
SDXL 0.9 release and features
Stable AI has announced the availability of SDXL 0.9, a major upgrade to the Stable Diffusion text-to-image model suite. Compared to its predecessor, the Stable Diffusion XL beta, the new version offers noticeable improvements in image and compositional detail. Currently accessible via ClipDrop, API coming soon. Research weights are available now, with an open release scheduled for mid-July.
Today, Stable AI Inc. announced the availability of SDXL 0.9, the state-of-the-art development of the Stable Diffusion text-to-image model suite. Following the successful release of Stable Diffusion XL beta in April, SDXL 0.9 greatly improves image and composition details over its predecessor.
Prompt: A wolf in Yosemite National Park, chilly nature documentary film photography
Negative prompt: 3d render, smooth, plastic, blurry, grainy, low-resolution, anime, deep-fried, oversaturated
Application scenarios and functions
SDXL 0.9 is capable of producing ultra-realistic images for film, television, music, instructional videos, as well as design and industrial uses. It also provides features such as image-to-image hinting, patching, and extensions.
SDXL 0.9 has great potential in creative applications, which can produce ultra-realistic works for movies, TV, music and teaching videos, etc., and also provides advanced technical support for design and industrial applications, making SDXL a practical application of AI images pioneer.
Prompt: ~aesthetic~~ manicured hand holding up a take-out coffee, pastel chilly dawn beach instagram film photography
Negative prompt: 3d render, smooth, plastic, blurry, grainy, low-resolution, anime
Technical details and advantages
The new model has a significantly higher parameter count, with a 3.5B parameter base model and a 6.6B parameter model integration pipeline. It uses two CLIP models, including the largest OpenCLIP model ever trained, OpenCLIP ViT-G/14. Despite the powerful output, SDXL 0.9 can run on modern consumer GPUs.
A key advantage of SDXL 0.9 is its substantially increased parameter count (the sum of all weights and biases in the neural network upon which the model is trained) over the beta version, which is a 3.5B parameter base model and a 6.6B parameter model integration pipeline.
Prompt: beautiful scenery nature glass bottle landscape, purple galaxy bottle (SDXL 0.9 - 1024x1024)
Feedback and Application
Since the beta release in April, nearly 7,000 users have used SDXL to generate more than 700,000 images. SDXL 0.9 is currently available on the Clipdrop by Stability AI platform, with a full open release, SDXL 1.0, planned for mid-July.
Since the SDXL beta release on April 13th, the response from our Discord user community (nearly 7,000 people) has been overwhelming. These users collectively generated more than 700,000 images, with an average of more than 20,000 images per day. Over 54,000 images have been submitted and 3,521 SDXL images have been nominated as winners in "Showdowns" in the Discord community.
Prompt: magical realism; manicured fingers holding a piece of white heart-shaped sea glass up against the setting sun realistic film photography (SDXL beta - 480x480)
How to get it and future plans
Currently, SDXL 0.9 is available on the Clipdrop by Stability AI platform. Stability AI API and DreamStudio customers will have access to the model on Monday, June 26, along with other leading image generation tools such as NightCafe.
SDXL 0.9 will be used for research purposes only for a limited period to gather feedback and refine the model before its public release. The code running on Github will be available to the public.
If researchers wish to access these models, please apply using the following links: SDXL-0.9-Base model, and SDXL-0.9-Refiner. Please log into your HuggingFace account with your academic email to request access. Note that SDXL 0.9 is currently limited to research use.
Scan code attention
Share, like, and watch are the biggest support