Stable Diffusion is the 2022 text-to-image AI model associated with Stability AI that generates images from written prompts.
Stable Diffusion was developed by researchers at CompVis, Stability AI, and Runway, with support from academic and industry collaborators. Its initial public release took place in August 2022. The model uses a latent diffusion process: it learns to remove noise in a compressed representation and can turn text guidance into a new image.
A major reason for its influence was that the model’s weights were made available under a license that allowed broad experimentation and local use, subject to its terms. This helped create a large ecosystem of interfaces, fine-tuned models, and creative tools.
Stable Diffusion is often compared with DALL-E, Midjourney, and Imagen. Those are separate systems developed by different organizations. The phrase “text-to-image” describes the task, not one particular model, so the model name matters when identifying the answer.