Stable Diffusion 3.5

Stable Diffusion 3.5 is a text-to-image model family for developers building image tools and creators producing visual work on their own infrastructure. Prompts can describe the subject and composition across photographic, painted, line-art and 3D styles.

• Model choices Large targets professional image work, Large Turbo favors faster generation, and Medium balances image quality with customization on consumer hardware.
• Visual guidance Large supports Blur, Canny and Depth ControlNets to guide generation with visual inputs rather than relying only on text.
• Application integration The Python reference library supports basic inference for organizations implementing the models, while ComfyUI provides another way to run them.

You can deploy the models in your own environment under a self-hosted model license. The reference code is MIT-licensed, but model weights are downloaded separately and carry a separate model license; the code license doesn't cover the weights.

Stability AI also offers remote access through its API, cloud partners and the browser-based Stable Assistant. Its image editing and upscaling services include object removal, inpainting, outpainting and background changes. Those services are separate from the image models you run yourself.

This listing covers SD 3.5 Large, Large Turbo and Medium. The Large model card identifies the Stability Community License for weights, with commercial eligibility conditions. The reference inference code uses MIT; that license does not replace the weight license.