In a groundbreaking announcement, the creators of Stable Diffusion 3, a cutting-edge text-to-image model, have introduced an early preview, showcasing substantial improvements in multi-subject prompts, image quality, and spelling capabilities. The model, not yet available to the public at large, has opened its waitlist for early access, providing enthusiasts an opportunity to experience its enhanced features firsthand.

Anime Artwork - Stable Diffusion 3

Early Access Preview – Join the Waitlist

Potential users are encouraged to join the waitlist for Stable Diffusion 3’s early preview, a pivotal phase for gathering insights to refine the model’s performance and ensure its safety before a widespread release. The waitlist sign-up is available here.

Versatility in Scalability and Quality

Stable Diffusion 3 offers a suite of models ranging from 800M to 8B parameters, aligning with the creators’ commitment to democratize access to advanced AI. This diverse range caters to users’ varied creative needs, combining a diffusion transformer architecture with flow matching. A detailed technical report on Stable Diffusion 3’s architecture will be published soon.

Model Parameters and Quality

Safety at the Forefront

Prioritizing safety and responsible AI practices, the creators have implemented several safeguards to prevent misuse by malicious actors. Collaboration with researchers, experts, and the community has been integral to enhancing safety measures throughout the model’s development, testing, and deployment.

Safety Measures

Democratizing Access to Generative AI

The commitment to making generative AI open, safe, and universally accessible is central to the creators’ mission. Stable Diffusion 3 aims to provide adaptable solutions empowering individuals, developers, and enterprises to unleash their creativity.

For those interested in exploring existing image models for commercial use, the Stability AI Membership page offers self-hosting options, while the Developer Platform provides access to the API.