Latest Developments and Public Release of Stable Diffusion

Recently, Stable Diffusion has announced its latest version — Stable Diffusion 3, which brings significant improvements in handling multi-topic prompts, image quality, and spell checking capabilities. While not widely available yet, developers have opened an early access waitlist aimed at further optimizing model performance and security through early user feedback.

Additionally, the public release of Stable Diffusion emphasizes efforts in security and ethics to ensure the responsible use of AI technology globally. This includes incorporating AI-based safety classifiers by default in the software package to automatically filter undesired outputs.

Stable Diffusion v2 is another notable version that utilizes an OpenCLIP ViT-H/14 text encoder and a diffusion model combined with an autoencoder to offer high-resolution image synthesis capabilities. This version’s model can generate outputs of 768x768 pixels and provides various classifier-guided conditioning ratios to optimize the quality and details of image generation.