Stable Diffusion Goes Public: An Open-Source Image AI Anyone Can Use

Stability AI publicly released “Stable Diffusion,” a text-to-image generation AI model. Rivaling DALL-E 2 and Imagen in capability, its defining feature was releasing the model weights themselves under the CreativeML OpenRAIL-M license — drawing enormous attention as an image generation AI that anyone could run on their own hardware.

Details

• Release format: The first public checkpoint was v1.4, with model weights published on Hugging Face
• Development team: A collaborative effort by Stability AI, RunwayML, the successor group to the University of Munich’s CompVis research group, EleutherAI, and LAION
• Prior rollout: On August 10, the model had already been given early access to around 1,000 researchers, with more than 10,000 beta testers already active on a Discord server
• Licensing: Adopted a permissive license allowing both commercial and non-commercial use, fueling rapid adoption within the open-source community

What happened next

The choice to go open source laid the groundwork for many later image generation AI services to build derivative models on top of Stable Diffusion. It marked the starting point of an evolution that continued with SD2.0 that November and SDXL the following year — establishing a “anyone can modify it” current distinct from DALL-E and Midjourney.