Stable Diffusion released as open weights
Stability AI and collaborators released Stable Diffusion, a latent diffusion text-to-image model small enough to run on consumer GPUs, with openly downloadable weights — democratizing image generation.
Key facts
- Public release 22 August 2022
- Based on 'High-Resolution Image Synthesis with Latent Diffusion Models' (arXiv 2112.10752)
- Trained on subsets of the LAION-5B dataset
- Ran on consumer GPUs with under 10GB VRAM
- Spawned a huge ecosystem (fine-tunes, ControlNet, LoRAs)
What happened
Anyone could download and run a state-of-the-art image generator locally, with a permissive license.
Why it matters
The open-weights release made generative AI a grassroots movement and set off legal battles over training data.
Changelog
- 2026-09-29: created
Related events
- DALL·E 2 brings photorealistic text-to-image generation ★★★★
- Ian Goodfellow introduces Generative Adversarial Networks (GANs) ★★★★
- OpenAI unveils DALL·E and CLIP ★★★★
Sources (3)
- officialStable Diffusion Public Release (Stability AI)
- paperHigh-Resolution Image Synthesis with Latent Diffusion Models (arXiv)
- codeCompVis/stable-diffusion (code)
id: 2022-08-22-stable-diffusion · updated 2026-09-29 · open in the interactive timeline