I started using DALLE-2 to see if there was any way I could use this technology within my web agency.
So far the results have been hit or miss. I've been able to generate some pretty cool images, but I've also been able to generate some pretty bad ones.
If you've seen the pictures that I generate for my blog, you'll know what I am talking about. Even so I found some utility for this tool and I wanted to share it with you.
In this example I wanted to see if I could use DALLE-2 or marketing purposes. I wanted to see if I could generate a picture of a can in a beach setting so I gave DALL-2 he following prompt,
"Can with condensation at the beach, next to palm trees and neon lights." The result was pretty spot on. The sand and condensation do look somewhat jagged, but even then it's better than not having the stock photos I am looking for. Besides I don't think I could have done a better job myself.
Even my inhouse designer was impressed with the results though it's still years away from being able to replace a human designer.
What really impressed me was how DALLE-2 as able to land the little details like the condensation on the can and the palm trees. The lighting was also spot on in my opinion.
Here I was having some fun and wanted to see if DALL-2 could be used to create some concept art for a video game. I used the following prompt, "Anime cyber samurai in futuristic city expresive."
I personally think this looks really cool despite the obvious flaws. The samurai's limbs seem to blend into the background. I still think that for something that was made in just a few seconds this is amazing.
At the very least an artist could use this as a starting point to get some ideas for a concept art piece. I had a hard time picking my favorite image from this batch. In general they all were very consistent with the prompt and
looked like something you would see in a behind the scenes video for a video game.
Here I wanted to see something colorful and I really like owls so I gave DALL-2 he following prompt, " Artistic owl flying through nebula colorful " I think this is one of the best images I have generated so far.
The colors are vibrant and the owl looks like it's flying through space. These type of images seem to be DALL-2's strong suit. I think it's because it's able to generate images that are more abstract and less realistic.
It also doesn't matter if the image is a little blurry or jagged because it's supposed to be abstract. This too can be used as a starting point for many type of artists. I can see future painters using prompts like these to
inspire themselves prior to creating a piece.
Lastly I wanted to test just how specific you could get with DALL-2 to create realistic images. I gave it the following prompt,"Bacon burger with cheese ingredients dropping on to steaming bun."
Even with the odd cheese squares I honestly couldn't tell the difference between this image and a real burger! I've seen plenty of small businesses that use bad stock photos for their menus and I think this could be a great alternative.
I could see this being used for menus in restaurants, grocery stores, and even fast food. The steamy bun looks like something straight out of a high budget commercial. Maybe that's because DALL-2 uses an enourmous amount of data to train itself so
I'm sure a lot of those commercials are used to create new content.
So is DALL-2 the future of AI? Honestly, I think it could be. I think it's a great tool for artists and designers. It's also a great tool for small businesses that don't have the budget to hire a professional photographer.
You won't get the same quality as a professional photographer, but you will get something that is good enough to use for your business needs. I think it's also a great tool for people who want to create their own art but don't know where to start.
I'll leave you with some of my favorite images that I have generated so far. I hope you enjoyed this tutorial and I hope you learned something new. If you have any questions or comments please feel free to send me a Tweet.
Swagger API documentation with NestJS to learn how to document and view the endpoints you just created. Feel free to reach out to me on Twitter @GmoMedel if you have any questions.
