Friday, January 20, 2023

AI Generated Comedians, Part 1

In recent months, I have been hearing a lot about computer programs that use artificial intelligence to generate art. I tried out five of the programs - NightCafe, Imgit, Stablecog, Midjourney and DeepAI. The programs require the user to type in a description of what they want to see. I figured to start out with something simple. I typed "Cheshire Cat." Here is what I received. 

Next, I typed "silent film comedian."

This was an odd result.

I decided to get more specific.  I typed "Buster Keaton."

Here is the result of "Buster Keaton on a bicycle". . . 

. . . and "Buster Keaton holding a cake." 

I was pleased with a program's rendering of Keaton as a child. 

Then, I tried something more generic - "comedian with a cake." 

How about "comedian sees a ghost"? 

Let's see what AI can do with "comedian chased by a dog."

Nightmarish, eh?

One of the programs had difficulty rendering Laurel and Hardy. 

I asked to see Clark Gable as Indiana Jones.

With Lensa, the user is meant to upload ten to twenty selfies of themselves. By using these images to understand what the user looks like, the program is able to generate creative avatars of the user. I did something different with the program. I uploaded portraits of several different silent film comedians. So, rather than generate an avatar of a single comedian, it mixed up the various images and created composite versions of the comedians.

No comments:

Post a Comment