Stable Diffusion Public Release logo

Stable Diffusion Public Release

Free

Announcement of the public release of Stable Diffusion, an AI-based image generation model trained on a broad internet scrape and licensed under a Creative ML OpenRAIL-M license. Stable Diffusion blog, 22 August, 2022.

FreeFree tier
Inputs: textOutputs: image
Type
Open Source
Company
Stability AI

About Stable Diffusion Public Release

Stable Diffusion is an open-source AI image generation model released by Stability AI in August 2022. It is trained on a broad internet scrape and released under the Creative ML OpenRAIL-M license, allowing commercial and non-commercial use. The release includes an AI-based safety classifier to filter unwanted outputs. The model requires approximately 6.9 GB of VRAM and is optimized for NVIDIA GPUs, with future support planned for AMD and Apple Silicon. The model weights (v1.4) and code are available on GitHub, and an optimized development notebook is provided for use with the HuggingFace diffusers library. A public demo and DreamStudio offer additional access.

Key Features

Released under Creative ML OpenRAIL-M license for commercial and non-commercial use
Includes AI-based safety classifier by default
Model weights v1.4, requiring 6.9 GB VRAM
Optimized for NVIDIA GPUs, with future support for AMD and Apple Silicon
Available via HuggingFace diffusers library with development notebook
Public demonstration space and DreamStudio for easy access
Community support via dedicated forums

Pros & Cons

Pros
  • Open-source and freely available for commercial and non-commercial use
  • Includes built-in safety classifier to filter inappropriate outputs
  • Strong community support and active development
  • Integration with popular libraries like HuggingFace diffusers
  • Multiple access points: direct weights, demo, and DreamStudio
Cons
  • Requires high GPU memory (6.9 GB VRAM) for full operation
  • Currently optimized only for NVIDIA GPUs; AMD and Apple Silicon support pending
  • Model may produce biased or unsafe content without proper safeguards
  • Large model size may be challenging for low-resource environments

Best For

Text-to-image generationCreative content creation for artists and designersResearch in AI-generated imageryPrototyping and concept visualization