Stable Diffusion is Stability AI's family of open-weight text-to-image models, and its defining trait is that it can be downloaded and self-hosted on consumer hardware rather than used only through a vendor's cloud service. The models are released under the Stability AI Community License, which is free to use — including for commercial purposes — for organizations with under $1 million in annual revenue.
Who builds it
Stability AI develops Stable Diffusion and distributes it in two ways: as open weights that anyone can download and run locally, and as a hosted service through the Stability AI API for users who prefer not to manage their own infrastructure, according to Stability AI's own announcements and stable-image product page.
What it does
Stable Diffusion generates images from text prompts across what Stability AI describes as virtually any visual style, including 3D, photography, painting and line art, according to its announcement of Stable Diffusion 3.5. Text rendering inside images is supported, per model documentation published on Hugging Face, and inpainting-style editing is available through the Stability AI API, according to the stable-image product page. The current generation, Stable Diffusion 3.5 Medium, generates up to roughly 2 megapixels of output, per Stability AI's own announcement.
Pricing
Stable Diffusion is open source: the model weights themselves are free, and Stability AI's licensing page confirms a free tier of use. The key qualifier is the Stability AI Community License, which governs when "free" includes commercial use. According to Stability AI's Community License Agreement and license page, use — including commercial use — is free for organizations with annual revenue under $1 million; organizations above that threshold need an enterprise license from Stability AI rather than relying on the free Community License terms. Stability AI's own materials do not publish a specific starting price for that enterprise tier or for its hosted API, so larger organizations should confirm current terms directly with Stability AI before budgeting a deployment.
Rights and provenance
Ownership of generated output sits with the user, according to Stability AI's Community License Agreement. Commercial use rights are real but conditional rather than unconditional: the same $1 million annual-revenue threshold that governs free use also governs free commercial use, which places Stable Diffusion in a "commercial with limits" category despite its open-source status — a nuance that is easy to miss given how often "open source" is read as "unrestricted." Because the models are open-weight, Stable Diffusion also differs structurally from closed, API-only competitors: an organization can self-host the entire pipeline, audit the weights, and run generation without sending prompts to a third-party server at all — a meaningful distinction for privacy-sensitive or offline use cases, though it also means there is no vendor-managed provenance watermark applied centrally in the way hosted, closed tools can enforce at the API layer.