On Twitter, independent designer Geoffrey Dorne embarked on a funny visual experiment. Throughout August, he illustrated press articles using the Midjourney artificial intelligence-assisted image generation tool. Mischievously titled LPPIin reference to the famous French press agency, this series of images allows, according to him, to illustrate information for which there are no photos or visuals. “A leak in a nuclear power plant, a lake with invisible pollution, or a beluga whale dying in a river, you can’t take a picture of them directly,” he explains. “As a result, AI seems interesting to me for showing things that we could not show otherwise. »
To obtain these results, Geoffrey mainly uses the titles and the beginnings of the articles in his prompts (the sentences from which the AIs make an image). The more the title is poetic or contains puns, the more unexpected and interesting the results. Casually, he probably just invented a new profession for the world of the press: “IA illustrator” .
When image synthesis becomes widespread
Our designer is far from the only one to have taken advantage of image generation tools in recent months. After the first wave of I.Artists who contributed to the making of these generators, a new generation of users is taking over. During the month of August, Steve Coulson, the creative director and owner of the writing assistance platform Campfire and photographer Kris Kashtanova have both published webcomics generated with Midjourney. In the press, too, things have changed. Media like Atlantic or Cosmopolitan generated their first illustrated front pages with Dall.E and MidJourney.
To understand this craze for AI-generated art, you have to go back a few months in the past. In April 2022, the Dall.E image generator created by the Open AI firm excited imaginations and created many frustrations. Although marketed, this tool remained closed to the general public. Only a technical and limited demo could be tested on the craiyon.com site. Since then, things have changed a lot. New open source and open to Internet competitors have appeared such as MidJourney and more recently Stable Diffusion.
Officially open since August 18, MidJourney offers image generation in exchange for subscriptions of 10 or 30 dollars per month. For its part, Stable Diffusion offers Going through its Dream Studio site more than 1,000 image creations for just under 12 euros. Better still, the most resourceful can run the code of this tool directly on a computer, completely free of charge. Besides the fact that the image generation Going through Stable Diffusion seems faster, the results are also close to what Dall.E 2 can give, which is still a benchmark.
Is that a good situation, AI illustrator?
This extension of the generation of images to as many people as possible is not, however, done without questioning. Seeing the first press covers illustrated with AI-generated images, artists have wondered if the tool will not supplant press photographers and illustrators.
For his part, Geoffrey Dorne has already had positive feedback from editorial staff from whom he has borrowed the articles to illustrate. ” Release is curious and rather pleasantly surprised by the result, he says. The Obs also wrote to me to say that it was interesting. I also had many people in the industry who were devastated and already saw the death of their profession. I think it’s an understandable reflex fear, however I’m not sure an AI can compete with a pro illustrator. An illustrator’s humor or culture on a subject is not yet something replaceable by a machine. As for the photographers, they are also taken for their look which is unique (not to mention that the AIs work with dated databases while the press photographers work on recent news subjects). AI illustration is bound to become mainstream, but it won’t kill other visual formats. »
The future of artistic creation in question
The world of the media is not the only one to wonder. Concept artists, responsible for imagining fantastic worlds in films and series, are beginning to wonder if their profession is going to die out. In a video posted on TikTok by speaker specializing in AI, Gianluca Mauro, an artist indicates that she already sees companies favoring image generation tools rather than hiring her.
“We laugh about it between us, but I have already lost two jobs because of Midjourney, explains the artist. I am encouraged to quickly learn how to write prompts and to become a specialist in this field. They are right but it really pisses me off […] The cool part of our job is that you have to think. There are so many parameters like colors, light, composition… These tools do it all for you and deliver it to you as is. There is no more thinking. »
The image that illustrates this article was generated by the Dream Studio tool and represents concept artists robots in front of drawing tables.