Stable Diffusion 3.5
A text-to-image model family you can deploy on your own infrastructure, with a consumer-hardware variant and separate API and web access.
Stable Diffusion 3.5 is a family of text-to-image models for people building image tools or producing visual work on their own infrastructure. It generates photography, paintings, line art and 3D-style images from prompts, with an emphasis on following the requested subject and composition.
The Large model is aimed at professional image work. Large Turbo favors faster generation, while Medium balances image quality with customization. Medium targets consumer hardware. For more directed results, Large can use Blur, Canny and Depth ControlNets to guide images from visual inputs.
You can run the models in your own environment with a self-hosted license. Stability AI also provides remote access through its API, cloud partners and the web-based Stable Assistant. Its image editing APIs cover tasks such as removing objects, extending an image and changing backgrounds; those services are separate from running the image models yourself.
The Python reference implementation is MIT-licensed and contains code for basic inference, including the text encoders, image decoder and diffusion model. Model weights are obtained separately under a model license. The reference code is intended for organizations implementing the models; ComfyUI is an alternative for running them.
