Stable Diffusion 3.5: A Creative Comeback!

The Story: Stability AI has launched its latest AI image model, Stable Diffusion 3.5, after the significant backlash from its predecessor due to its inconsistent outputs. This new version aims to remedy past issues with improved performance, diversity in image generation, and models that can be easily run on consumer hardware, embracing community feedback and encouraging creative freedom.

The Details:

• Stable Diffusion 3.5 introduces multiple model variants: Large, Large Turbo, and Medium, with support for customization and accessibility.

• The Large model features 8 billion parameters, allowing users to generate high-quality images up to 1 megapixel, while the Turbo version ensures faster output at the cost of some quality.

• Customized outputs will now better highlight diversity, generating images with different skin tones without extensive prompting, reflecting a commitment to inclusiveness.

• The Medium model, which launched on October 29, is optimized for consumer hardware, capable of generating images from 0.25 to 2 megapixels, catering to casual users and small businesses alike.

• Stability AI emphasizes ownership and permissiveness, allowing creators to monetize their generated media while keeping them free for smaller businesses.

Why It Matters: With Stable Diffusion 3.5, Stability AI is addressing user frustrations and embracing a more agile approach to development.