I made this blog with the intention for this year to become more creative. I’m a big advocate of CGP Grey’s new year’s theme and mine this year is “create more than you consume”. It encapsulates my desire to be more minimalist and creative. In my effort to flex my creativity muscles, once I heard about Dall-E a few months ago I had to try it out.

For those of you who aren’t familiar with Dall-E, it’s a prompt-based image generator. You give it a prompt and Dall-E will spit back a few images inspired by that prompt. For example, when I looked up “Arnold Schwarzenegger in ukiyo-e style”, I got this:

Arnold Schwarzenegger in ukiyo-e style

Pretty cool right? You can also upload images to use as inspiration to create similar images. Using previous Dall-E images in an iterative process can really help you to get exactly what you’re looking for.

This was the original image I used: “Impressionist, monk meditating on top of a mountain with a cityscape underneath”
I really like the addition of the smoke in this one.
I like that the city looks more vibrant in this one.
This one is my all-time favourite Dall-E image I made.

As you can tell from the prompt I used to make the first monk image, it really pays to be specific when using Dall-E. There’s a bit of a learning curve when it comes to crafting the right prompt and I still continue to falter when using Dall-E. There’s a balance when it comes to being specific, creative, and out of the box and understandable by the text-processing software. The corpus of Dall-E’s image inspiration also doesn’t contain everything I’d like to satisfy my artistic appetite.

“Japanese pagoda in the style of Derain” Although Dall-E missed the mark on incorporating the Fauvist art aesthetic I was looking for, I really like the images it made for me, particularly this one. The colourful clouds in the background remind me of those drawn on Tibetan thangkas.
Why does he have robes like Doctor Strange?
Keanu looks more like a generic Jesus in this one
Don’t look too closely at Keanu’s face or else you’ll deep dive into the uncanny valley
“Keanu Reeves in thangka style” I was trying to essentially get a bodhisattva thangka with Keanu’s face. Unfortunately, it doesn’t seem like Dall-E really knows what a thangka is. However, it’s clear it knows it’s supposed to be religious and vaguely pattern-y. I think this one is the closest to the mark when it comes to Keanu’s face and the thangka imagery.

One consequence of trying to craft the perfect prompt is that I’ve really had to read up on my art history to get a precise answer. It doesn’t always work but I’ve found it to be helpful at times to hone in on an aesthetic style. Rainbow blotches? Afremov. Little dot painting? Pointilist. I’ve also been looking at more art to get inspiration for more Dall-E images.

Despite how cool Dall-E is, it does have several limitations. First, I’ve found that sometimes the lines aren’t very defined; I’ve stuck with Impressionist and watercolour styles to get around this. Second, Dall-E sucks with making human faces. It can nail the general features but it can’t do the fine details. If you look closely, all the humans Dall-E makes fall deep inside the uncanny valley. Third, Dall-E can’t do text. It just puts in random gobbledygook or illegible handwriting. It’s a shame too because I’ve been also really into reading about fonts and calligraphy, especially different Asian scripts. Fourth, Dall-E’s language processing just isn’t 100% there; I don’t really expect it to know what a thangka is but I was trying to make a cartoon image of a World War II shark mouth plane and I couldn’t get it to understand I want to plane with shark art not a plane that is a shark. Finally, even though generating images through an iterative process can be helpful, it’s usually focused on the wrong aspects more often than not wasting my precious few credits I have for Dall-E.

Red Asian dragon with a cyberpunk cityscape in the background. The lines aren’t super clear.
Falconer holding eagle on the Mongolian steppe. Don’t stare at the Mongolian guy’s face or he’ll eat your soul. The eagle looks kind of funky too.
“Tibetan flames in minimalist style” I was trying to get the flames found on heruka thangka
Robot samurai bowing down cyberpunk realistic
“Thich Quang Duc, minimalist tattoo style” Clearly Dall-E doesn’t know who Thich Quang Duc is
World War 2 shark plane, comic book style. Very creative, Dall-E, but not exactly what I was looking for.
“Robot samurai bowing down cyberpunk realistic”This was honestly 90% there to what I wanted but I was hoping he’d be more samurai looking
Iterative from the left image. Dall-E leaned more into the robot than the samurai and for some reason made it a robot cowboy instead? Still looks really cool though

Final thoughts

People worry about the rise of AI replacing everthing we do and Dall-E seems to be a step closer to the inevitable robot dystopia. Text-based image processing has had meteoric improvements in the past two years and it’s amazing that I can use Dall-E to create artwork a whole lot faster and better than anything I could make.

But I don’t see Dall-E as replacing human creativity. Instead, I see it as a complement to it. For people like me who aren’t good at art (yet), Dall-E lowers the barriers to entry and allows them to flex their atrophied creative muscles. Coming up with a cool concept is a task in itself and to have a piece of artwork generated from that idea in mere seconds is a leap in making art more accessible. I’ve learnt so much art and art history since making my account to get ideas for prompts or the language to express the aesthetic I’m looking for. I actually like art museums now because Dall-E has given me a better appreciation for art and beauty!

Even for the experienced artist, Dall-E can be helpful for inspiring artwork. You don’t have to spend hours and hours seeing if something looks good when you can input it on Dall-E and voila! I’d imagine the randomness of Dall-E image generation can help to change your perspective on a concept.

More than replacing human work, technology supplements it. I don’t think that Dall-E cheapens art but deepens it by making it more democratic to newbies and less tedious for artists. Personally, it’s absolutely turned me into a more creative person and I even want to try painting the monk meditating into the urban abyss image sometime. If you haven’t already, I highly recommend trying out Dall-E and experiencing it for yourself.

Here are some more of my favourite creations: