Models
Stable Diffusion
An open-weight latent diffusion model released by Stability AI in 2022; first widely available text-to-image generator runnable on consumer GPUs.
Stable Diffusion is a text-to-image diffusion model first released in August 2022 by Stability AI in collaboration with researchers from CompVis (LMU Munich) and Runway. It applies the latent diffusion technique introduced in the 2022 paper High-Resolution Image Synthesis with Latent Diffusion Models, performing the iterative denoising in a compressed latent space rather than raw pixels — which is why it can run on a single consumer GPU.
The original 1.x versions, then SD 2.x, SDXL (2023), SD 3 (2024), and Stable Diffusion 3.5 progressively improved image fidelity, prompt adherence, and resolution. Crucially, the weights were released under a permissive licence, enabling an enormous open-source ecosystem: ControlNet for spatial conditioning, LoRA fine-tunes for style and characters, IP-Adapter for image references, AnimateDiff for video, and front-ends like AUTOMATIC1111, ComfyUI, InvokeAI, and Fooocus.
Stable Diffusion’s open release reshaped the AI image landscape, made on-device generation a real category, and triggered ongoing debates about training-data sourcing, deepfakes, and copyright. Together with closed competitors such as Midjourney and DALL-E, it defines the contemporary state of AI imagery.