The first people to draw comics with AI have already made money

Produced | Tiger Sniff Youth Culture Group

Author | Mu Zitong

Editing, cartography丨Za County

Born in the Cambrian explosion era of AI, every day when you open your eyes, you have the illusion that you will be laid off tomorrow.

However, before AI snatched the job of human beings, it was still cunning humans who first found the way to get rich by squeezing AI.

Recently, a new form of comics is refreshing the perception of the publishing world:

Humans are responsible for creating stories, and AI is responsible for the artists—never-before-seen "crash comics" can also be published for money.

For 1,280 yen, for the price of a tonkatsu set-up, I bought "Cyberpunk Momotaro".

On the cover, an Asian teenager in oil style gazes out of the picture, his pink hair resembling neon on an iron-gray background.

A hundred years from now, the comic book could be put in a museum, along with Apple's original machine, marking some important points in the digital age.

Now, it is Japan's first officially published AI drawing manga monobook, which has just been released on March 9.

There are 147 pages in total, and they smell the ink that has just come off the printing press.

The reason why it is an AI drawing manga and not an AI manga is because "Cyberpunk Momotaro" is not entirely created by AI.

All images are by AI, and the story script and storyboard are provided by humans.

Its human author is Rootport, a science fiction comic book writer who loves to try early adopters, and the AI author is Midjourney, a popular image-generating artificial intelligence that just released V5.

Initially, Rootport just posted it as a game on Twitter.

Unexpectedly, the goddess of traffic immediately patronized, it was like an electric music divine comedy sitting on a rocket, spread wildly on the Internet uncontrollably, and quickly attracted the attention of the publishing house.

Rootport's Twitter account is serializing a new AI drawing comic

The story begins with a philosophical line:

"Have you ever imagined that if after falling into a deep sleep for 5 minutes every night, the universe ends, and every morning 5 minutes before waking up, the universe is reborn. So, what does 'death' mean? ”

Along with this line, a teenager staggers into the night of the slums, and behind him, the cheesy neon blurs into a dirty palette.

He is the protagonist Momotaro, an unknown male corpse that should have been disposed of by a cleaning robot, but is accidentally picked up by his husband and wife who are doing illegal human body business.

Since it is called "Momotaro", it is naturally still a classic story of Momotaro, but this time the stage is set in the background of cyberpunk in the near future.

In the cyberpunk "New Okayama", humans replace their original brains with electronic brains, collectively completing the "computerization" of the brain.

The tech giant ONI who mastered electronic brain technology became the de facto ruler, the gap between the rich and the poor in society widened dramatically, and countless new gray products were bred in the slums on the edge of the city.

Here, an old couple lives.

They are like traditional Momotaro folk tales, living by chopping wood and raccoon, respectively.

However, the husband chopped electronic firewood and used hacking technology every day to exploit the loopholes of the government and large enterprises.

The old woman washed banknotes, and she laundered local ash by running an underground strip brewery.

In order to facilitate money laundering, the old women used to buy illegal "brain dead" human bodies, which were re-plugged into the electronic brain and programmed by the husband to use them as cheap striptease dolls.

But one day, the transport cabin transporting the "living corpse" suddenly arrived early, and in it lay not a human body suitable for making dancing girls, but a mysterious boy with pink hair.

He lost all his memories and was like a newborn baby.

But contrary to his impoverished brain, he also carries a treasure of great value.

It is the legendary ONI's most advanced secret weapon, "KBY", a brand new program that is said to have control over the world.

Why is such an important treasure in the boy? Who is he?

WITH THE POWER OF THE TEENAGER, WILL THE OLD COUPLE BE ABLE TO RECOVER THEIR DAUGHTER WHO DISAPPEARED IN SEARCH OF KBY?

The old couple took in the amnesiac teenager and named him "Momotaro", and a legendary story began.

While the backstory sounds exciting, Cyberpunk Momotaro is far from a full-fledged manga.

There is no problem with the story, there is no problem with the script, the problem is still in the picture that the AI is responsible for.

The whole comic is like a strange dream of imaginary patients, which will collapse if you are not careful.

The protagonist's face is completely uncontrolled, and in the previous frame, he may be a stubborn street boy with freckles, and the next frame is a K-pop boy with a big nose.

Distinguishing characters depends on the "identification anchor" set by the author:

The white-bearded black man is the husband, the silver-haired white man is the old woman, the pink-haired is Momotaro, and the wolf-eared girl is the police "puppy".

For less demanding readers, this style of drawing does not affect understanding.

Like a story book with a word order upside down, it seems to be gibberish, but in fact, the brain can automatically correct it during the reading process.

But for a comic, such drawing skills are undoubtedly failing.

As some readers complained:

"Such a good world view setting, such good lines, it's a pity that the AI ruined the painting, and the characters' faces changed back and forth, which was too difficult to read. If I could come up with a cartoonist's handmade version, I'd definitely like to read it again. ”

In Japan, the country of manga, people scold AI paintings for having no soul while being moved by "Cyberpunk Momotaro".

What kid hasn't dreamed of being a cartoonist and drawing everyday whims into small picture books? It's just limited by the handicap, and the ambition is stillborn, leaving a few turtle graffiti to lead the neck in the math book.

Now, all you need is tapping on the keyboard and providing a few hints to get a free and super-fast drawing assistant.

Compared with the advantages of AI crash comics, those flaws in drawing are less fatal.

"I want to try it too!"

At the end of the comic book, Rootport reveals his comic book production process.

Compared to a traditional comic book hand-drawn workshop, this is more like a tango dance between man and AI.

First identify the story and characters, then create the screen in Midjourney.

Enter keywords such as "pink hair, Asian boy, cyberpunk, motorcycle jacket, manga style" and you'll get a series of portraits of Momotaro.

Choose the one you like best, enlarge it, refine it, add new cues to it that express different emotions, generate new images, and accumulate them in the library.

Then prepare the storyboard, and the AI will generate the corresponding image according to the picture required by the storyboard.

Finally, the image drawn by the AI is edited into the page through the comic software, and it becomes a complete comic.

During the whole process, what Rootport does is provide prompt words, adjust prompt words, and the most time-consuming and labor-intensive drawing work is all done by AI.

Imagine that the comic is like a delicious pizza, with human authors sprinkled with all kinds of delicious condiments, and AI artists incorporating these condiments into pizza pies to bake a mouth-watering work.

Analogous to the traditional Japanese manga production process, you can think of AI as the original art assistant hired by the author.

It's just that this assistant is too genius, so he doesn't listen to the command very much.

Rootport started working on Cyberpunk Momotaro last year, when he used a V3 version of Midjourney.

As we all know, AI had several classic problems at that time.

First, I can't draw my hands, just like an art examiner who always wanders in class, when it comes to the drawing hand, it is either multiple fingers or a mess.

The second is that there is no guarantee of continuity in generating characters, and using the same prompts, such as "blonde", will generate a completely different face each time.

Third, the overall control of the image is weak, and drawing the face alone may perform well, but if you draw a full-body image, it is easy to cause the face to collapse.

Therefore, Rootport can only constantly adjust the "spell" that drives the AI to circumvent the problem.

He calls this process the "gacha machine": you are prepared for everything you can do, have a rough idea of what the AI will generate, and then leave it to fate.

When it really can't be adjusted, Rootport is very comfortable choosing to put it on the table:

Technology can only do this, and I have done the limit of what I can do, and I believe that a faceless nude does not affect the reader's understanding of the plot.

Cyberpunk Momotaro is far from perfect, but it shows a future direction.

Is it possible for a person who can't draw but can tell a story to become a cartoonist?

Now it's possible.

From idea to completion, a 110-page full-color comic can be completed in less than a month with the help of an AI assistant.

Even if you don't have a day of art education, you can become a crash cartoonist.

However, at the same time, an age-old question was also put on the table -

Does this AI-generated crash comic deserve copyright?

A clear answer to this question has already been given by countries.

On March 16, the U.S. Copyright Office announced that AI-generated works are not protected by copyright law.

The previously copyrighted American comic book Zarya of the Dawn was also reclaimed.

The comic, also generated by Midjourney, is the work of Kristina Kashtanova and made a splash last year after successfully applying for copyright protection.

Unexpectedly, two days after being happy, the copyright office changed its mind again, pointing out that there were no human-written elements in the works output by Midjourney, so it could not obtain copyright protection.

However, in Japan, as we have seen, express manga is still under the umbrella of copyright law.

That's how they understand it - a quick comic is about the same as a script writer hiring a drawing assistant.

In the Japanese manga world, this mode of creation has long been a regular operation.

For example, "Warrior of Love" Hyun Fuchi, as a screenwriter with a handicapped hand, often contacts cartoonists with suitable drawing styles after writing the script and asks the other party to create comics based on the script.

The story king Aki Takagi in "Dream Eater" and the manga master Majo Takami are also such a pair.

Even professional cartoonists who are good at drawing will hire some drawing assistants.

Because Zhou Geng's comic production cycle was too tight, many well-known cartoonists themselves completed storyboards and important picture line drafts, and shadows, backgrounds, and coloring were handed over to assistants.

Therefore, which chicken laid the egg is not the focus of Japanese readers, the focus is still whether the egg tastes good or not.

After Zarya of the Dawn was decopyrighted, author Kashtanova filed a similar defense.

Her lawyer pointed out that although Midjourney automatically generates images, it generates pictures because of the cartoonist's prompt, after which the cartoonist also selects the picture to form a comic story, which contains a lot of "creativity unique to humans."

Just as a picture taken by a monkey with a camera cannot be copyrighted, so can a human.

What matters is not what tools are used, but who is using them.

In the end, the debate is still about the "human content" in AI-assisted works.

In response, the U.S. Copyright Office countered in the subsequent "Guidelines for Copyright Application of AI-Generated Works":

Although many people believe that the provision of "cue words" by humans represents human intervention in the creative process, in terms of results, the output of Midjourney is uncontrollable by humans, so it cannot be counted as human creation.

However, this seems to make sense of analysis, and with the recent release of Midjourney V5, it has gradually become a bit untenable.

After the Midjourney V5 optimization algorithm, the output of the prompt word becomes more stable and controllable, which is undoubtedly challenging the theoretical foothold of the US Copyright Office.

The upper and lower figures generate underground strip bar scenes for V3 and V5 respectively, and you can see the improvement in accuracy

And this is only the change of the world in half a year, no one can say, at what speed AI painting will continue to evolve next.

When AI-generated images can be as accurate as PS's brush, how can copyright law distinguish between the contribution of people and tools?

The 2020 AI manga "PHAEDO" was generated using the original manuscript of Osamu Tezuka as a training library and did not encounter copyright challenges

The AI copyright dilemma we face today is like a classic thought experiment in Cyberpunk Momotaro.

If someone now invents a mechanical neuron, and you replace one neuron in the brain with a mechanical neuron, is that brain still your original brain?

The answer is, of course, yes, one out of tens of millions of neurons is replaced, and the impact is insufficient.

So what if you go further, replace a tenth of neurons?

This should still be your brain.

What's next, replace half and replace all?

Is your mind, memory and soul still the same as it was?

This is the scary thing about AI, which blurs the boundaries between intelligence and thinking, and makes us have to face questions that have been asked for thousands of years, but have never been so urgent -

What exactly is a "person"?

How should "human content" be defined?