DALL-E 2 (also DALL·E 2) is a deep learning model by OpenAI where you can generate digital images based on descriptions in natural language. The original version, DALL-E, was first mentioned in 2021 and introduced in 2022.

The text and image embeddings are from another OpenAI network called CLIP (Contrastive Language-Image Pre-training). It finds the best caption for a picture as input. The goal of CLIP is to understand the relationship between an object's visual and textual representations.

DALL-E 2 uses over 10 billion parameter training versions of the GPT-3 transformer model and is trained on millions of stock images, which makes it especially helpful for creating images for corporate use.