Why Dall-E will not steal my job as an illustrator
I finally was able to try it. And I can assure you it’s impossible to create professional and useful images from a text description. Here are my tests trying to create art with Dall-E.
I am a professional illustrator and storyboard artist since 10 years. I work mainly for advertising, and a little bit for cinema. When I started to read all those articles about Dall-E and Midjourney, saying that AI will steal my job as an illustrator, I was very scared. Then I became more skeptical when I understood how it worked. So I registered on their waiting list, and I finally got the opportunity to try it this morning, to see if I could recreate something similar to what I do for a living, on it.
I simply tried to recreate professional images that I had to do for a client, especially a simple storyboard frame for Citroën car company. I totally randomly choose that frame, that was made for an instagram ad (that’s why it’s vertical). Here is the original frame:
As you can see, it seems very “easy” to describe. It’s a man and a woman, in a Citroen Ami car, front view. I don’t even mind that the car is colored in blue and the characters are white. I just wanted to see if I could get something similar with Dall-E. The results are better than I expected, but really far from being useful for my client. Here a a few of the results I got with different descriptions:
The software didn’t know the exact model of the car (because it’s new probably). The characters are ok but not so realistic. And getting the same composition was almost impossible. I also realized that I forgot to mention so many details that seemed intuitive for me, like having the characters smile, looking in front of them, having the woman driving, the characters having their seatbelts on etc…
So all of that to say that it’s very, like very, difficult to describe an image with words. There are so many things that an illustrator will instinctively know how to do, that the machine would need to be told to do. Even if it was possible, it would take a very long time to get a description coherent and satisfying for a client, and many more time to try many tests, to get something right. It took me maybe 10 to 15mins to generate all those images (thinking about the correct descriptions but also generating the images took a few minutes). None of them are really usable, in a professional presentation I mean. And it took me something like 30mins to draw it myself, including the sketch to validate with my client the correct angle and composition, and correcting with their feedbacks.
Now imagine having to do this process again and again, and again, to finally get a coherent storyboard of 12 images like this one, with the same characters and cars in different angles:
So yes I have big doubts that those softwares, even vastly improved, will really help my clients in a professional mindset. It could help maybe to create fast moodboards. But I doubt they would want to save some money to get something like that. They need the drawings to be good, to be precise, to look professionals and to communicate exactly what they want. They are doing extremely expensive advertising campaigns (in the range of a few hundred thousand dollars, even millions when it’s for TV), so I doubt they would want to save a few hundreds (or even thousand) dollars by replacing their storyboard artist with a software. They need everything to be professional and “on point”. To explain the idea correctly to their boss, their clients, the movie director, the camera team, the special effects department etc…
A storyboard is a professional tool, to have a coherent vision of a project. It can’t be imprecise, approximate. And I’m not even talking about having the same characters with different and strong expressions, like on this storyboard I did for Brico Depot:
It’s a little bit like when camera became cheap, and everybody had a good camera on their phones, people assumed it would be the death of professional photographers. But it was not. People still hire professional photographers for their wedding (like my sister recently), because it looks so much better when done by a pro. And advertising agencies and Fashion brand companies still pay large sum of money to professional photographers to shoot their campaign. I doubt any serious company will just say “let’s just buy a good camera and shoot the campaign ourselves, it will be cheaper than hiring a real photographer!”. A tool is just a tool. You still need someone specialized in using it, and with a good artistic sense.
I actually took some photography classes during my studies. I had a very good camera. And I can assure you, taking good photos is very very difficult. I admit I’m a bad photographer. It’s just not my art field. If I had to make professional photos, I’ll hire a real photographer. It’s not because you have a good tool that you are a good artist.
We could say the same thing with painters. When photography was invented, people assumed paintings were dead. They are not. When digital art was invented, traditional art continued to strive. Those are just different art forms, with their own specialists.
Talking about art forms, how Dall-E is doing with purely “artistic drawings”? I tried to recreate an illustration I did, not in a professional context, but just for myself. Because then there is less incentives to be “precise” (as it’s not professional). I just wanted to see if Dall-E was better for art than for professional storyboards. So here is the image I had made myself (I’m a big fan of Vaporwave and lofi aestethics, I do a lot of those kind of 80’s images for my instagram, then I animate them):
And here are the images that Dall-E made, with my description of this image:
It’s like, I don’t really know. It’s just not what I had in mind. Of course, because it’s almost impossible to explain a mental image with words. Also, evidently, it’s not my “style”. It totally lacks personality. But I mean, to be fair, it’s still impressive for a machine. It probably will get better over time. But do you understand what I mean, when I say that we still have many many years before people can create interesting art using those tools?
Then I tried something different. I uploaded my image (above), and asked Dall-E to create variations of this image, just to see. Here are the results:
Yeah… Take a closer look to one of them:
I think you see the problem haha.
And what if the softwares become way better in the futur you may ask?
As I said on a previous article, even chess survived the success of AI tools. When Deep Blue beat Garry Kasparov in 1997, everybody assumed chess was dead. But more and more people are learning and playing the game today than before. So many youtube and twitch channels dedicated to chess have millions of subscribers. And the world champions like Magnus Carlsen are still making millions of dollars.
At the same time, no human player is able to compete with the computers. They actually use the computer to learn and better understand the game.
So why chess is not dead? Because in sport like in art, humans are more impressed and interested by humans abilities, even when they are no match for other machines or animals. Almost nobody watches matches between chess computers, it’s a little bit boring.
Also, today, with a gun, you can kill the strongest fighter man on earth in one second. The most trained martial artist, who spent 30 years crafting his art, can be defeated by a child who found a gun. We could assume that martial art should be dead. It’s “useless” compared with technology. But we still love to watch fights as a sport, between opponents fighting with their hands. Because it’s not the ultimate result (being stronger or beating your opponent) that is important, it’s the human excellence that people love to watch and admire. It’s not impressive to kill someone with a gun. Very much more to watch someone doing it with their hands, against an equally trained fighter.
A machine can lift a house, but people still love to participate in and watch the strongmen competitions.
A car, or even a cheetah, can easily outrun Usain Bolt, but we are still impressed by him, because he is the fastest human. We don’t care if any average Joe is faster on his motorbike. We are impressed that a man can RUN faster that anybody else. Not that he IS faster than anybody or anything.
So yeah, even if Ai becomes better at creating illustrations than humans (and I still have some doubts, I’m not saying it’s impossible, I’m just skeptical), I still believe my field will survive, like chess.
The cool thing will maybe not anymore to see the final results, but to see people in the act of drawing. A little bit like today, some automated pianos can play anything, but people still love to see real humans play piano. It’s more fulfilling to see humans do and create things, than to see a machine doing it.
So to all the people talking about the end of illustration and animation, or even art, be aware that art cannot die. It’s an idea. Ideas don’t die as V said. Moreover, it’s an idea central to human life since the beginning of times. Since a guy, in a cave, 50 000 years ago, realized he could put a print of his hand on the cave wall, using dirt, blood or anything. Or that he or she could sing, dance, tell stories…
Ai is a tool. What makes art is not a tool, it’s the will to make art. A machine doesn’t have will. A tool can transform art, improve it, expand it, not erase it. I’m very excited to see what those tools will bring to futur artists. I’m not worried they will steal our jobs. I mean ok, maybe just a little bit deep down haha, but not in the short or mid term at least, that I’m sure of :)
Have a good day!