Which 2022 text-to-image AI model was released with open-source code and weights?

Stable Diffusion was the 2022 text-to-image AI model released with open-source code and weights.

Stability AI, CompVis, and Runway announced Stable Diffusion in August 2022. The model generates images from text prompts by using a diffusion process: it learns to reverse controlled noise additions until an image matching the conditioning information emerges.

Unlike many earlier image generators that were available mainly through a hosted service, Stable Diffusion could be downloaded and run by users with suitable hardware. That distribution model helped trigger a rapid expansion of community tools, interfaces, fine-tuned models, and creative experiments.

The phrase “open source” is sometimes used loosely in discussions of the model. Its code and model weights were made broadly available, but the exact licensing and training-data questions remain more complicated than the phrase suggests. Stable Diffusion is also distinct from DALL-E 2 and Midjourney, which were separate image-generation systems.