Stability AI has launched Stable Diffusion 3.5, featuring two new models—Stable Diffusion 3.5 Large and Stable Diffusion 3.5 Large Turbo—following a less successful release of SD3 Medium. This iteration aims to enhance visual media transformation based on community feedback. The models are designed for accessibility on consumer-grade systems, promising improved image quality and rapid generation capabilities, with the Large Turbo variant producing images in about 40 seconds. The new licensing model permits free use for small businesses earning under $1 million, contrasting with competitors like Black Forest Labs. Key features include three versions tailored for different user needs, improved customizability through Query-Key Normalization, and support for negative prompts to refine output. Additionally, SD3.5 rivals leading models like Flux in areas of prompt adherence and resource efficiency. While it offers uncensored creative freedom, users may find variability in less specific prompts. Stability AI is expected to roll out Control Nets by October 29, adding advanced features to the workflow, suggesting they are forging a robust path toward future development in AI-driven image generation.

Source 🔗