When OpenAI presented their Text-AI GPT-3 the experts, users and researchers were thrilled. Rarely has one seen such an impressive use of AI that can write text on its own with very little information. Now, the research network is stepping it up a notch and releasing DALL-E, an image-AI that is just as amazing as its text-based relative already was.

In DALL-E, a pun on the name of famed artist Salvador Dalí, OpenAI brings together the capabilities of its text generator GPT-3 with the image generation capabilities of another project, i-GPT, to create something never before seen.

DALL-E can generate matching images using text input, creating not only extremely arbitrary combinations, such as a radish in a tutu, but also complex illustrations and motifs.


DALL-E in action (source: Mixed)


DALL-E can work with real and with fictional elements and create paintings based on a few lines of text, which are almost like artists like Dalí himself. Also existing pictures can be adapted by DALL-E with the appropriate text input and even be changed by the perspective.

The Image-AI in action (source: Mixed)


Like its predecessor, the new Image-AI uses contextual information to reinterpret and transform existing information. When the system, to which so far only some sample videos exist, will be released or whether it, similar to GPT-3, will soon be available to first alpha and beta users, is still unclear.


Post picture: OpenAI


Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.