Stability AI announced Stable Diffusion 3, its new text-to-image AI model version

Stability AI has launched Stable Diffusion 3, an enhanced new version of its previous text-to-image models like Stable Diffusion. The new version offers improved interpretation of multi-subject prompts, superior image quality, and better spelling capabilities, with parameters ranging from 800 million to 8 billion.

The model employs a diffusion transformer architecture and flow matching techniques to deliver top-notch results, especially when it comes to better typography implementation than prior Stable Diffusion models, putting it in competition with established models like OpenAI's DALL-E 3 or Midjourney. The new model also addresses the shortcomings of its predecessor, such as the accurate interpretation of multi-subject prompts and coherent spelling, making it more versatile across different applications.

Users interested in this can join the waitlist on the Stability AI website for early access and to receive exclusive benefits like access to a dedicated Discord server for updates.