Generative Art

I tested Generative Art within DALL·E 2. The results are as followed; I choose the best example of what I wanted but I will say the GA was limited.

Comment

Inspiration

I wanted to see how much an AI art builder can make with detailed descriptions, stylized art, artistic terminology, and basic art understanding.

What it does

The DALL·E 2 generated images from machine learning to create the image I requested. However, the ML failed to understand what I wanted and gave me many blank results; with a term, I seemingly asked too much of the machine. I asked for a person in the flower field; it failed to generate one.

How we built it

It is a free generator (requires your data through third-party logins) that I did not build.

Challenges we ran into

The ML didn't understand full descriptions and often fell short of completing an image I wanted.

Accomplishments that we're proud of

I would say that some aspects impressed me; the castle and cyberpunk models--along with different art styles (vaporwave, digital art, surrealism, etc.). Other styles like pop art weren't working, and basic add-ons wouldn't compute.

What we learned

These AI art generators aren't the best; room for improvement. But they do have the capability to create good art in a particular style. The stylized art isn't where you'd assume it would be for an AI/ ML. Especially since minor adjustments make the machine freeze.

What's next for Generative Art

Not sure. I thought the ML would be advanced with the ability to assume (personalization) that humans would want art with different examples (significant changes, not minor ones) within the four options given. These AI/ MLs require millions of artistic input; to become more efficient the creator of these machines would have to pay artists--something that may not happen.

Built With

  • dall?e
Share this project:

Updates