Stability AI has just released SDXL 0.9, which is the newest version of their Stable Diffusion text-to-image model. This updated model brings major improvements to the quality and detail of the generated images compared to the previous version. It is a powerful tool that can be used to create highly realistic images in various creative industries.
SDXL 0.9 brings exciting new features that allow users to create incredibly realistic visuals for films, TV shows, music, educational videos, design projects, and industrial applications. The model’s impressive abilities open up a world of possibilities for using AI-generated images in practical situations.
The SDXL series goes beyond basic text prompting and offers users a variety of features.
With SDXL 0.9, users can now experiment with the following features
- Image-to-image prompting
- Inpainting
- Outpainting
These functions provide more creative options and flexibility in generating different versions of images, filling in missing parts, and smoothly extending existing images.
In simple terms, it allows users to be more creative and make changes to images in more advanced interesting ways.
The Power Behind Stability AI SDXL 0.9
This impressive improvement in image quality achieved by SDXL 0.9 is due to a significant increase in its parameter count. The base model of SDXL 0.9 has 3.5 billion parameters, and when combined with the ensemble pipeline model, it reaches a total of huge 6.6 billion parameters.
This makes SDXL 0.9 one of the image models with the highest number of parameters available as open source. The pipeline uses two CLIP models, including the powerful OpenCLIP ViT-G/14, which boosts processing power and enables the creation of highly realistic images with a resolution of 1024×1024.
Accessible to Modern Consumer GPUs
SDXL 0.9 can be used by people who have a modern consumer GPU, like the ones found in regular computers. The requirements are not too demanding.
You need to have the following:
- Windows 10 or 11, or a Linux operating system.
- At least 16GB of RAM
- Nvidia GeForce RTX 20 graphics card
- Graphics Card with at least 8GB of VRAM
If you use Linux, you can also use a compatible AMD card with 16GB of VRAM.
Positive Reception during Beta Launch
Since it was introduced as a beta version on April 13, SDXL has received an incredibly positive response from its Discord community, which now has almost 7,000 users. In a short period of time, users have created over 700,000 images, averaging more than 20,000 images per day.
The community has also actively participated in the ‘Showdowns’ competition, submitting more than 54,000 images. Out of these, 3,521 images generated by SDXL were chosen as winners. The enthusiastic engagement of the community demonstrates the popularity and success of SDXL in generating high-quality images.
Availability and Future Development
Starting from June 26th, users of the Clipdrop platform by Stability AI, as well as customers of Stability AI API and DreamStudio, will be able to access SDXL 0.9. This means they can take advantage of its advanced features.
Additionally, popular image generating tools like NightCafe will also offer access to SDXL 0.9. In the near future, Stability AI has plans to release SDXL 1.0, which will improve the model even more and provide additional functionalities. This shows their commitment to refining the technology and making it even better.
Promoting Research and Collaboration
Stability AI will allow researchers to use SDXL 0.9 for a limited time and gather feedback to improve the model before it becomes widely available. Researchers can request access to SDXL-0.9-Base and SDXL-0.9-Refiner by logging into their HuggingFace Account with their academic email. It’s important to note that SDXL 0.9 is currently only for research purposes.
SDXL 0.9 is an innovative tool that enables industries to explore creative possibilities using AI-generated images. Stability AI is working on SDXL 1.0, which will further enhance the capabilities of the model and solidify their position as a leader in the field of AI-assisted creativity.
Also Read: Stability AI is Leading the Way in Large Language Model Development
Conclusion
Stability AI SDXL 0.9 is a significant advancement in the field of generative AI imagery. The model’s improved image quality and composition detail make it a valuable tool for a variety of creative and industrial applications.
As the model continues to develop, it is likely to have an even greater impact on the way we create and interact with images. We hope you found this article helpful. If you have any questions, please feel free to leave a comment below.
