Playing around with DALL·E 2

Earlier today, I got access to play around with OpenAI’s DALL·E 2, it’s a system to generate images in almost anything imaginable by taking in text input and generating a few variations of it. The more descriptive your prompt for the art, the more likely it will get the details.

I’m going to use this post to share some prompts that yielded me interesting results.

My first prompt was:

a dachshund chasing a squirrel as an impressionist painting

I wanted to try something very specific and unlikely to be found on the internet.

an old woman walking through a graveyard in a village in pakistan, digital art

Another one with a Pakistani scene

a busy bazaar on a hot summer day in pakistan with merchants selling fruit on a cart, oil painting

It’s not very good at faces of people or animals, so things as impressionist or non-realistic art can look great. Right when you try to generate a face as the main subject, it has a hard time:

majestic lion looking over the Serengeti, photograph

I’m going to try more things in my spare time but these were some initial prompts I tried.