Stable Diffusion is Stability AI's open-source text-to-image model. See how it works, pricing, versions, and key features for creators and developers.
Stable Diffusion is a text-to-image generative AI model developed by Stability AI, a UK-based AI company founded in 2019 by Emad Mostaque. First released publicly in August 2022, Stable Diffusion generates original images from text prompts using a diffusion-based deep learning process.
What set Stable Diffusion apart from competitors like DALL-E and Midjourney was its open release of model weights, allowing anyone to download and run the model on their own hardware, including consumer GPUs, rather than only through a hosted application. This openness spurred a large ecosystem of community tools, fine-tuned model variants, and third-party interfaces such as Automatic1111 and ComfyUI.
Stable Diffusion generates images from natural-language text prompts and also supports image-to-image transformation, inpainting, outpainting, and fine-tuning on custom datasets. The latest major release, Stable Diffusion 3.5, introduced Large, Large Turbo, and Medium model variants that trade off image quality against generation speed and hardware requirements, and it can run on consumer hardware rather than requiring specialized data-center GPUs. Because the model weights are open, developers can fine-tune Stable Diffusion for specific art styles, product categories, or brand aesthetics, and integrate it directly into their own applications rather than depending on a third-party hosted API.
Stable Diffusion's model weights are free to download and run locally under the Stability AI Community License, which permits free research, non-commercial, and commercial use for organizations or individuals with under 1 million US dollars in annual revenue; organizations above that threshold need an Enterprise License from Stability AI. For users who prefer not to run the model themselves, Stability AI's hosted API uses a pay-as-you-go credit system priced at about 0.01 US dollars per credit, with no monthly subscription required; a single image from the Large model costs roughly 6.5 credits, and the faster Turbo variant costs about 4 credits per image.
Yes, the model weights are free to download and run under the Stability AI Community License for research, non-commercial, and most commercial use; large enterprises need an Enterprise License.
Stable Diffusion was created by Stability AI, a company founded in 2019 by Emad Mostaque and headquartered in London.
Stable Diffusion was first released publicly in August 2022.
Stable Diffusion 3.5 is the latest major model family, released in October 2024, including Large, Large Turbo, and Medium variants.
Yes, because its weights are open, Stable Diffusion can run locally on consumer GPUs using tools like Automatic1111 or ComfyUI.
The API is pay-as-you-go, priced at about 0.01 US dollars per credit, with a Large model image costing roughly 6.5 credits.
Its weights are openly available under the Stability AI Community License, though the license includes usage conditions rather than being a fully unrestricted open-source license.
It is used for AI art generation, product and marketing imagery, game and concept art, and as a base model for custom fine-tuned image generation tools.