About DALL-E

Let's play with these upcoming models...

DALL·E is a Machine Learning (ML) model trained to generate images from text descriptions. DALL-E is trained using a dataset of image and annotated text pairs downloaded from the internet. Then, the model can be fed a string of descriptive text, and it outputs a constructed image best representing the text annotation.

The first version was released by OpenAI in January 2021. A subsequent improved model capable of generating images of higher resolution called DALL-E 2 was released in April 2022. In a separate exercise, DALL-E mini was created as an open-source model to mimic the results of DALL-E, but provide a smaller, easy to use model, and without restrictions on commercial usage imposed by Open.AI. DALL-E Mini is available for exploration at Craiyon.com. DALL-E is available for use via select invitations.

The beauty of DALL-E models is that they have been proven adept at generating images which may not be realistic. The classic example that DALL-E authors introduced was generating im…