DALL-E was OpenAI’s 2021 text-to-image model that generated images from written descriptions.
OpenAI announced DALL-E in January 2021. The model was designed to create images from natural-language prompts, including unusual combinations of objects, styles, and settings. Its name combines the artist Salvador Dalí with the animated robot WALL-E.
DALL-E was based on a version of the GPT-3 architecture adapted to model text and image tokens together. It demonstrated that a single generative system could connect language with visual concepts, although its output quality and consistency were more limited than those of later image generators.
DALL-E is often confused with DALL-E 2, released in 2022, and DALL-E 3, released in 2023. Those later systems improved image quality, prompt following, and text rendering, but they are distinct generations rather than alternate names for the original 2021 model.