Testing out DALL-E 2

21 June 2022

I was recently granted access to DALL-E 2, the impressive new AI system that can generate images from text prompts.

I don’t have anything technical to say about it, I just wanted to share some of the incredible images it has produced. If you’d like to know more about it here are some links;

Anytime you enter a text prompt and click generate you are presented with six versions of your request, see below for a close up of a pizza.

Here is another example of the six different images Dalle-2 generates for each request you submit.

You can specify the style of the image, or it will make an assumption about which style might be more appropriate. For the hippos above, I did not specify that they should be a cartoon, but Dall-e made assumed that a cartoon style would be the best approach. Similarly for a man walking in the street in Khayelitsha, South Africa it chose a very photo realistic approach.