DALL-E, named after the artist Salvador Dalí and Pixar’s WALL-E, is an AI that can generate images from text prompts using datasets of text-image pairings. It can do a variety of tasks such as drawing human-like animals and objects, creating images using two unrelated concepts, illustrating text, or altering pre-existing images. Even though DALL-E can generate many unique images, it is far from perfect. 

OpenAI’s DALL-E generates images based off of a text prompt. The results and success rate can change depending on how the prompt has been phrased, or how many objects or details it has. For example, let’s say the prompt is “a hedgehog wearing a red hat, yellow gloves, blue shirt, and green pants.” DALL-E would not only have to place the correct clothing items on the hedgehog, but it would also have to match the items to their assigned colors. Adding more items and colors to the prompt can take away from the quality of the final picture. This also applies to its ability to correctly position or stack objects and control various features within an image.

Although DALL-E’s technology is already pretty cool, there’s still a lot of space for improvement. If you would like to get more information on DALL-E, you can read more on OpenAI’s official website.

RELATED STORIES:
OpenAI’s new machine learning AI model generates images from text