London-based artificial intelligence (AI) laboratory Stability AI has unveiled an early preview of its latest text-to-image model, Stable Diffusion 3. The advanced generative AI model is designed to generate high-quality images from textual prompts, boasting enhanced performance in critical areas. This announcement closely follows the introduction of Sora, a similar AI model by Stability AI’s competitor OpenAI, capable of producing nearly-realistic, high-definition videos from basic text prompts.
Stable Diffusion 3, while not yet available to the general public, showcases notable improvements in handling multi-subject image generation compared to its predecessors. This enhancement allows users to incorporate more intricate prompts with multiple elements, resulting in superior outcomes. The model also claims advancements in overall image quality and spelling accuracy, addressing historical issues of consistency and coherence found in previous text-to-image models.
Although Stable Diffusion 3 is currently in a private preview stage, Stability AI has initiated a waitlist for individuals keen on early access. This preview phase enables Stability AI to gather valuable user feedback and refine the model before a broader release slated for later this year.
Stability AI emphasizes its commitment to safety and responsible AI practices, stating that it is working with experts to assess and mitigate potential risks associated with Stable Diffusion 3—mirroring OpenAI’s approach with Sora.
“In preparation for this early preview, we’ve introduced numerous safeguards. By continually collaborating with researchers, experts, and our community, we expect to innovate further with integrity as we approach the model’s public release,” said Stability AI.
Stable Diffusion 3 is offered in a variety of model sizes, ranging from 800 million to 8 billion parameters. Stability AI aims to strike a balance between creative performance and accessibility for users with varying computational resources.
The laboratory remains committed to ensuring that generative AI is open, safe, and universally accessible, aligning with its mission to unlock human creativity. “With Stable Diffusion 3, we strive to offer adaptable solutions that enable individuals, developers, and enterprises to unleash their creativity, aligning with our mission to activate humanity’s potential,” stated Stability AI.
