geeker
I had already written about Wombo.art in detail. The AI of this software generates partly fascinating, but mostly very vague artwork from keywords, which lives more from the mood and the atmosphere than from what it actually shows. While I find this very groundbreaking, various readers loudly disagreed.
For a few weeks now, AI algorithms have been available on the net, which once again put a significant shovel on it – and whose possible effects have to be discussed slowly. Shit is getting serious.
It’s hard to deny: Powerful tools are growing here, not only for book covers and posters – the entire field of commercial graphics is facing a radical change if you just have to throw a few keywords at the AI so that it can provide professional-looking illustrations of almost any style.
The “hot shit of the day” is MidJourney, an AI that is fed via a Discord channel. If you are not familiar with the topic, it sounds confusing at first, but in the end it is comparatively simple:
I came up with the topic today because author and graphic designer Michael Marrak brought it up on Facebook. He is certainly affected: if authors can design convincing covers of their books themselves, graphic designers become obsolete.
Is this still a dream of the future or a reality that is just not yet accepted?
You can really let yourself be shot away on the showcase page of MidJourney, where particularly magnificent creations are exhibited. If you click on the pictures, you will even find out according to which keywords they were created.
To be clear: These are not collages from Internet sources, no “mashups” – these are originals created by the AI in 60 seconds.
The fact that MidJourney sees itself as an open community has advantages and disadvantages, if I have understood correctly: you can actually use the generated images – with minor restrictions also commercially – but they do not become the property of the person who generated them. Everything belongs to everyone.
The first steps can be done for free, then a subscription is necessary, which I find absolutely fair priced: 10 dollars a month for 200 requests to the AI.
However, it takes a little practice to create expressive graphics. My entry into MidJourney was rather bumpy yesterday. Nevertheless, the results impress me and I would like to share them with you.
Maybe one or the other Wortvogel veteran remembers that I once wrote a few Nibelungen novels and am very impressed by the artwork, even if they are “only” digital in nature. So it was obvious to feed MidJourney first with “barbaric warrior fights dragons in the style of a painting”.
Unfortunately, I mistyped, made the warrior a plural and the dragons are of course not fixed in quantity. Nevertheless:
It’s pretty feisty, dynamic and atmospheric. It helps that you can ask the AI to take some of the four suggestions again as a basis for four more suggestions. And the favorite can then be extrapolated into decent resolution (maximum 1664 x 1664 pixels). Strictly speaking, this is not yet ready for printing, but programs such as Pixelmator Pro also use AI to scale it properly.
However, MidJourney has some very curious limitations – I wanted to generate the image of a “seductive Marilyn Monroe type floating through the clouds” and had to be taught that “seductive” is a blocked term and the use of blocked terms can lead to a ban on the user. They seem to want to oppose Rule 34 at all costs.
My next idea: A bizarre, disturbing motif where I leave a lot of room for MidJourney: “Elvis Presley as an alien”. If I had planned to write a satire about an alien who is making a career on Earth as a rock ‘n’ roll musician, then this would always be an excellent cover for it:
Let’s poach a little more in the territory of Messrs. Marrak, Thiemeyer and Eckartsberg: suppose I want to set a novel in Manhattan, which has been destroyed and reconquered by nature. How would MidJourney illustrate this? AI’s answer:
Anyone who now interjects “there are no people to be seen” – that’s true. But unlike Wombo, it’s only because I didn’t ask MidJourney to do so.
Now it all looks like oil paintings and this style is really not suitable for every project. MidJourney can also be different. The request to generate laughing teddy bears with the attribute “realistic” can trigger nightmares:
Of course, these are all just gimmicks with no real value. What could I really do with such an AI? Next experiment. Years ago, I was working on a fictional horror novel called “A Ticket to the Funeral Procession,” in which a subway appears late at night in London, controlled by skeletons and demons, bringing people to hell (a tribute to “The Train to Hell,” one of my favorite stories by Robert Bloch).
So quickly typed “old London Underground in the station, populated by skeletons and demons” into the command line. Result:
For a “ghost thriller”, this would not only be perfectly fine, but miles better than what the publishers are currently doing.
If you like it futuristic and are more into science fiction than horror – how about an armed cyborg ninja in a synthwave metropolis?
Since we have furnished our apartment in MidCentury/Palm Springs style and are still looking for a little matching artwork, I type “Downtown Palm Springs at sunset in bold colors”:
That’s pretty much what we want. And I get an idea. I have been friends with Lee Goldberg for several years, who is now the driving force behind the very successful crime publishing house/network BRASH BOOKS. Their covers are designed by professionals and look like this:
That’s what I want! So fix over to PosterMyWall, template selected, artwork uploaded and adapted:
Maybe I have to say this clearly again: for this pulp crime cover, which is absolutely marketable in this form, I didn’t need two minutes. Cost: 0 Euro.
Maybe that’s also the reason why there are more and more klautor:innen – because all the trappings of a book (design, printing, marketing) are getting easier and cheaper and you only see the actual content as something that you then have to get “somehow”. As if the novel were not the core of the book, but just one of the many random pieces of the puzzle.
But I digress.
When trying to design a crime cover, the current limitations of MidJourney are once again clearly shown: Words like “blood” and “bloody” do not let the AI through and the attempt to generate a gun bleeding from the barrel on a table in a dimly lit office leads to disastrous results – unlike a graphic designer, MidJourney has no idea, what is meant. The AI already fails in its attempt to represent a pistol:
There is no denying that AI does not replace a graphic designer with whom you can consult, who understands intentions instead of keywords and who can make the idea “behind” the image understandable.
But that doesn’t change the fact that MidJourney is an entirely new way to visualize ideas. All you need is imagination – no longer mastery of tools. The sky is the limit – or is it?
To quote A CACTUS OF ALL THINGS IS NOT A LUTSCHBONBON:
“Perfection? What is that? Reactionary-fascist outgrowth of stubborn diligence and talent. Art is politics and that means: bringing the less fit into line with those who determine what art is.”
Of course, working in a Discord channel with a command line is not very comfortable. I suspect it’s only a matter of time before this is outsourced to its own app or at least to a clearer web interface. I also miss a few functions: you should be able to adjust individual keywords afterwards without having to start a completely new AI query. It would also be ideal if keywords could be weighted so as not to leave the distribution entirely to AI. And finally: ideally, it should be possible to upload certain elements (an existing figure, a logo, a lettering) in order to integrate them seamlessly into the artwork.
In addition to MidJourney, there has also been a lot about DALL in recent months. E 2 have been reported, but I have simply not dealt with it sufficiently. However, this video gives a good introduction to the topic and shows successful examples of the capabilities of the algo:
With this, we have clarified WHAT the new graphics AI can do. But what about the effects? Will graphic designers soon be sitting in the pedestrian zone wearing their hats and begging for alms? Will AI replace an entire profession? Will it be able to interactively simulate a conversation in which the user specifies his wishes? Is this a brave new world or a nightmare?
As always, the answer is (presumably) in the middle. As impressive as MidJourney’s graphics are when you see them out of context, the AI still struggles to hit exactly what the user needs. But this can also be an advantage, because MidJourney’s proposals also contain the potential for variations and previously unimagined alternatives.
Professionals and companies will certainly not want to do without having their products designed by professionals and companies for the next few years.
But that’s just a postponement, and when you look at the evolution of AI creativity over the past decade, it’s hard to imagine where we’ll be in another ten years. A good example is the series STAR TREK: DISCOVERY, which is already written by AI bots that have been trained with technobabble and newspeak, but which lack any human understanding of characters, stories or plausible interactions. There is still a long way to go.
One area in which algoes like MidJourney can already exert a massive influence today are the Low Budgtet productions, which so far could not afford good graphic designers, be they publishers, film companies or game manufacturers. They are able to use AI to at least professionalize the visual appearance of their offer. Something like this no longer has to be:
The artwork is just one of the many pages on which AI is currently attacking the creatives of the world. Composers, actors, authors, translators – they all have no reason to sit back in their armchairs and think “I’m glad I didn’t become a graphic designer”. You are not safe.
But that’s a topic for another day.
And with that, as always, I hand over to you. How do you see that? Does it tingle in your fingers to play with MidJourney? Is this a trend like Tamagotchis, which nobody will talk about in two years? Or is it just the tip of the iceberg of upheavals in which technology will replace creative work in the medium term after physical work?
GIPHY App Key not set. Please check settings