DALL-E 2
OpenAI's second image generator, the first to bring photorealistic text-to-image to the mainstream
DALL-E 2 is OpenAI’s second text-to-image model, announced in April 2022. Unlike the original DALL-E, it switched from an autoregressive transformer to a diffusion-based architecture, using roughly 3.5 billion parameters for the core model plus another 1.5 billion for the upsamplers that sharpen output resolution. It generated images at four times the resolution of its predecessor and introduced inpainting and outpainting, letting users edit parts of an existing image or extend it beyond its original borders. DALL-E 2’s viral demos of photorealistic and stylized images helped push text-to-image generation into mainstream awareness well before ChatGPT existed, and it opened to the public in 2022 after an initial waitlist period. OpenAI has since folded image generation into DALL-E 3 and its GPT-image models, and DALL-E 2 remains available mainly for legacy API access.