Stability AI has unveiled its latest artificial intelligence image generation model, Stable Diffusion 3.5, highlighting the ability for users to have greater customization options and support for running on local hardware. The model comes in 3 variants:
Stable Diffusion 3.5 Large 8B parameters, capable of producing high-resolution images at 1 megapixel
Stable Diffusion 3.5 Large Turbo, a customized version of Large focusing on faster performance with comparable image quality
Stable Diffusion 3.5 Medium 2.5B parameters, improved training for running on user devices with image resolution ranging from 0.25 to 2 megapixels
Stability emphasizes that users can easily adjust image details in Stable Diffusion 3.5, suitable for integration with applications and optimized to run on standard user hardware without heavy resource requirements. Particularly, the Large Turbo and Medium models produce more diverse image results without the need for manual detail adjustments.
Testing against competing models, Stable Diffusion 3.5 Large excels in image quality and accuracy, while Large Turbo complements speed and performs comparably with other models. The Medium version also outperforms similar-sized models in quality.
The Stable Diffusion 3.5 Large and Stable Diffusion 3.5 Large Turbo models are available for download through Hugging Face and GitHub under Stability AI Community licenses, free for general use and commercial usage revenue below 1 million dollars per year. The Stable Diffusion 3.5 Medium model will be released on October 29th.
Source: Stability AI
TLDR: Stability AI introduces the cutting-edge Stable Diffusion 3.5 AI image generation model with enhanced customization, hardware support, and superior image quality across its three variants. Available for download with licensing conditions, these models offer a balance of performance and quality for various user needs.
Leave a Comment