DALL·E 2
OpenAI · April 2022
☠ discontinuedCloseddiffusionimage
Parameters3.5B
Why It Matters
Made AI image generation mainstream — its photorealistic outputs sparked widespread public fascination and debate about AI creativity.
Description
Switched from DALL·E's approach to a diffusion model (which generates images by gradually refining random noise), guided by CLIP (a model that understands the relationship between text and images). Produced much higher resolution, more photorealistic results with the ability to edit existing images.
Notable Milestones
- ▸First widely available AI image generator to the public
- ▸Sparked global conversation about AI art and copyright
Key Innovations
Diffusion
DiffusionGenerates outputs by gradually denoising random noise into coherent images/audio. The backbone of Stable Diffusion and DALL·E.
Text-to-Image
Text-to-ImageGenerating images from text descriptions — the technology behind DALL·E, Midjourney, and Stable Diffusion.
Related Research (1)
DALL·EDiffusion
2021 · OpenAI
Demonstrated that a single model could generate diverse, creative images from arbitrary text descriptions, combining language understanding with image…