I’m not going to get into the habit of posting AI art pictures, but it’s been really fun seeing what I can come up with, and I have a feeling that as time goes on, the fact that so much AI art breaks copyright will cause significant changes in what publicly available AI art generation engines like Dall-E will be able to create.
It’s a lot of fun to play around with various prompts to see what you can come up with. There is definitely an art to crafting a prompt to get images looking the way you want them to. I’ve become relatively adept at tweaking the prompt to place various elements where I want to within an image. It’s not always apparent when looking at an AI generated image what’s going on behind the scenes.
It’s also great how you can generate a quick image to meet a unique need; for example, to accompany a blog post, create a meme, respond to a comment on social media, or illustrate an example of something. I’ve already whipped up images for a variety of purposes. That said, for the most part, creating a proper image takes many, many generations. For one good generation there might be five or six generations that miss the mark in one way or another.
If nothing else, it will be interesting to see where this technology goes in the future. Many of these pictures are amazing, especially when viewed from a distance. But often, when you make the image larger, or zoom into specific areas within the image, the reality breaks down in a serious way. Human faces and especially eyes seem to be one of the major issues (for example, in the below images of Alice and Shaggy). Words and text are also an issue; when you prompt the generator to include a sign (for example, “S-Mart” in the below example of Ash from Evil Dead), many times more text will also be included that contains words taken from the prompt, but misspelled or misrepresented.
As this technology improves it will be incredible to see what we’ll be able to do with it. I’m looking forward to it!
I started by creating some fun depictions of some of the monorails that frequent Walt Disney World:
Monorail Blue at Magic Kingdom.Monorail Yellow at Disney’s Polynesian Village Resort.Monorail Green.Monorail Red at Epcot.
And I continued by creating whatever visions came to mind:
An image of Chell from Portal in the world of Valheim. I feel the mood of this one turned out particularly well. The prompt for this one was quite complicated. I am currently using this image as my Steam profile picture.I input a description of the Mountains of Moria, copied straight out of Tolkien’s Lord of the Rings, to see what the AI image generator would come up with. I created a few more of these; it was fun to put images to the actual text in the books.I created this image in response to the current trend of posts on LinkedIn, where many users are currently “open to work” due to numerous layoffs that have happened recently, especially in the technology sector.A meme I created (but did post, because it’s way too niche) with Ash from Army of Darkness. A classic movie! This was an example of where prompting the generator to add text resulted in strange unwanted text. All I asked for was a large sign reading “S-Mart”, but with this came various signs with different misspelled examples of “peas porridge”, which was also included in the prompt but not as part of the text.Sorry for the political nature of this, but it was an image that came to my mind.In response to my friend Peter’s post about his plans to create a gingerbread house for the Christmas holiday season.An A-frame cottage at a lake in Ontario’s cottage country.An image of a steam train and passengers in the style of Norman Rockwell.Alice and the Cheshire Cat in Manhattan. This is an example of an image where the eyes just aren’t generated right.Scooby Doo and Shaggy running from the hamburger monster. In this example it’s interesting to see how the hamburger monster’s arms and legs were generated using Scooby Doo’s legs and paws as a guide; putting “the hamburger has arms and legs” into the AI prompt without any further direction seems to have resulted in its using the other elements added to the prompt as inspiration. Every image generated in this series was the same way.This is an image of a wullamedian, a monster I drew when I was maybe 10 years old. It was basically a large cone with legs, a big circular mouth, and googly eyes on eye stalks. This is the closest representation I could create using AI. This is one of the images that took many generations to create, yet I still wasn’t able to fully replicate what I was envisioning.
I ended by creating a few Pomeranian images:
A Pomeranian at Barnard College in Manhattan.Pom fu.
I’m looking forward to creating more of these, especially as time goes on and the generation engines continue to improve. It is certainly entertaining!
