the development of the profession of neuroartist
Hi, my name is Alina and I am a neuroartist. In the TG channel “Parallel editions“I post my works generated in a neural network. You can treat the previous phrase differently: with disdain (“there can't be YOUR OWN works from a neural network”), with indignation (“I also draw in a neural network, but I don't claim to be a neuroartist”), with interest (“Wow, I'll read it, because I can't draw what I want”) and even with awe (“I still haven't gotten around to drawing anything in a neural network, but the person has already become a neuroartist”). Before arguing about the right to be called a neuroartist, it's worth first answering a more general question: can the generation of images in a neural network be called art?
There is nothing new under the sun: discussions about art have been going on for centuries. During the Renaissance, there was a debate between realists and idealists about what came first – ideas or the visible world – and the Dutch Minor, masters of genre painting, were criticized because their paintings did not correspond to ideals and often depicted everyday life. In the 19th century, photographs were considered mechanical reproductions of reality, devoid of soul and creativity. In the first decades of the 20th century, films were considered entertainment rather than art, until The Birth of a Nation (1915) and Battleship Potemkin (1925) were released.
In this article, using digital art as an example, I will break down the meaning of an image into its component parts and reassemble it, showing how we can complicate an illustration layer by layer using a neural network. With each new layer of information superimposed, the brain will read the image as more interesting and profound. In addition to conveying my views on neuroart, the article is of practical use. I will provide prompts, that is, text queries for generating images, in one of the neural networks – Midjourney – using the example of a water lily flower. To use the prompts for your needs, you will need to replace the words “water lily” with any other object.
What is art
Using the example of cinema, it is easiest to explain what kind of work we will call art: one in which the viewer sees several layers.
A romantic evening, lovers are sitting at a table by candlelight, the hero says to the heroine: “I love you.” And if by these words he means that he loves her, then this is not art: when the author describes exactly what is happening, and the characters say exactly what they think, this is only the first layer, “the expressed” – the ideas and feelings that the hero decides to reveal to others.
In a good movie, the viewer will hear the second layer, the “unspoken” – thoughts and feelings that the hero says in an inner voice and only to himself. “I’m not sure how you’ll take these words and what they mean to you,” “Please, just don’t push me away,” or even “I hope you’ll fall for the hook.”
The third layer, the “unspoken” – unconscious desires and passions that the hero cannot express in words even to himself, because they are too vague and difficult to define. “I need to be with someone, I can’t be alone”, “When no one pays attention to me, I want to die” or even “I have to pursue unavailable women to feel confident”.
Sometimes there may be a fourth layer – behind all the hero's experiences, you can see the position of the director, who already knows what is destined for the heroes: shared happiness or separation.
If you saw the difference between the presented images and felt that different shots of a couple in love carry different content, then you will be interested in reading this article to the end.
The Place of the Neuroartist in Fine Arts
Digital art is a form of art created using computer technology or digital devices. It can include various forms such as computer graphics, digital painting, animation, multimedia art, digital photography, and even interactive installations. Digital artists use specialized programs and tools to create their works.
A neuroartist creates a digital image, but only in a certain way – by turning text into an image in a neural network. However, the main difference between neuroartists and all other masters is not the uniqueness of working with a neural network: anyone who can type can learn to create an image from text. I see several characteristic features of the work of a neuroartist:
A wide arsenal of techniques and tools accumulated throughout the history of mankind. To use it, you must constantly expand your horizons.
The ability to train the neural network and improve its skills – and not alone. The technique of execution is outside the artist and all users of the neural network work on improving it.
From the previous point it follows that you can never get the same image from one prompt. They will be similar, but always different, and over time even change.
As the arsenal and horizons expand, the neuroartist is able to improve his visual skills quite quickly compared to other creative people.
Below I will consider three meanings that I put into the concept of digital art and provide prompts for generating each. Then I will combine the prompts to show that the image becomes more interesting and complex, you want to look at it, you want to think about it and draw conclusions. I will give examples of using such images in articles on IT topics. And finally, I will draw a conclusion about whether all this can be considered art.
The void from which the image is born
It’s hard for me to imagine the void that preceded the Big Bang. However, even when I was far from programming, I understood that in the virtual world, an object can be created from nothing and surrounded by nothing. Deep darkness, concealing billions of possibilities — this is how I imagine fertile soil for growing digital flowers. Here are some ways to depict the void.
The particles that make up an image
I can't even see a molecule or a cell without a microscope, let alone atoms and quarks, so the physical world seems fairly complete in my mind. However, when I think of the digital world, I imagine Neo from The Matrix, surrounded by green numbers, and old computer games with huge pixels. A characteristic feature of a virtual image is its decomposition into cells, and not all cells can be filled. A virtual flower does not necessarily start from a root: it can also start from a petal.
By the way, the ASCII art I linked to in this prompt is a good example of how the title in the prompt doesn't match the result. This image looks more like Minecraft than esky art, but the result suits my purposes, so I'm using this prompt.
The screen on which the image is projected
The first two points, emptiness and particles, can also be applied to a painting on a sheet of paper. So what makes digital art different? It is designed and created to be displayed on a screen. On a screen there can be noise, grain, unnatural colors, glare, different contrast and brightness, the screen can break. The water lily originates from the depths, but blooms on the surface of the pond. Display is the final layer of meaning that I see in digital art.
Two layers
Now I will show the effects that combinations of different prompts from the three groups above give. Some prompts have proven to be particularly stable and predictable, so I will attach several successful images to them, not just one.
Emptiness + Particles
By combining emptiness and particles in one painting, I can guess the process of how a water lily flower is created literally before my eyes and is about to take shape.
Particles + Screen
At precise coordinates, the pixels on the screen — in the process of transmitting signals from thin-film transistors to liquid crystal panels — reflect the results of the work: a flower emerging on the border of the real and virtual worlds. The image in these paintings is more static, although it seems as if it is about to crumble.
Emptiness + screen
There is nothing about pixels and digital code in this combination of prompts: a complete picture hits the screen straight from the void. A digital flower that jumped to the surface without going through the process of creation and therefore has not yet formed, leaving a feeling of incompleteness and great speed.
Three layers
The combination of three styles in the promts attracts the eye and looks more complex than all the previous paintings.
Neural network training
All the images above would be impossible to create if tens of thousands of people from all over the world did not generate images in Midjourney according to their ideas of beauty. As a neuroartist, I often train the neural network: I select the best images, enlarge them, create variations and do this again and again until the neural network learns to create the expected result for a certain prompt.
Here, for example, is my promt, which I spent several evenings learning six months ago:
Since then, the neural network has evolved and the combination of Nick Veasey and dotted pattern projecting texture produces amazing results.
Here's another prompt. A few months ago, I trained a neural network to create a magnolia flower during the design process:
Promt is still not very stable, although I managed to achieve a similar result for the water lily:
By the way, what will happen if I use all 15 styles I mentioned in one prompt?
As you can see, there is nothing special about this image compared to the previous ones. Prompt writing is not an area where quantity turns into quality.
Practical application of promts
René Magritte has a painting called “The Treachery of Images,” which depicts a smoking pipe with the inscription “This is not a pipe,” because it is not the pipe itself, but its image. Here is what Magritte himself said: “This famous pipe. How people have reproached me for it! And yet, can you fill it with tobacco? No, it is only an image, isn’t it? So if I wrote “This is a pipe” under the painting, I would be lying!”
In the prompts above, I am not teaching how to create digital art, but showing how to create an image that would indicate that this art is digital. Such images can be used in articles and advertisements to describe digital products, emphasizing their creation, content, or display. Below, I will give examples based on the same 15 styles that I indicated above, although there are several thousand of them. In my work, I have used more than a thousand styles.
For example, if I want to talk about how our company created a logistics system for a transportation company, I will highlight the design process in the image.
If I talk about the characteristics of a system, I will show it as robust and tangible, without taking away the focus from the fact that it is a digital product. That is how I would emphasize the reliability of the system under any conditions.
And that's how it is – speed.
And with the help of the following illustration I would hint that the competitors’ developments are not of high quality.
Let's take another example – a dating app. If users care about finding a partner based on compatibility analysis, I would focus on the algorithms behind it.
If users are looking for one-time dates, I would add color for passion.
If users are looking for long-term relationships and marriage, I would add an extra layer of old-fashionedness.
Or maybe someone is expecting a love story like in the movies?
Author's view
I have tried to describe how a neuroartist can achieve the desired impression depending on the context, but this is still not art, but visualizations and advertising images. So what will make neuroimages art? Exactly the same thing that makes a painting, a photograph or a movie art. It is the ability of the artist to remove the brush from the canvas when the painting is finished; the ability of the photographer to press the shutter at the right moment; the ability of the director to frame the shot and start shooting. The combination of technical mastery and creative intuition is called the author's vision.
The author's view is distinguished by the fact that its effect on the viewer is impossible to predict: the picture lives its own life, it has not three or four meanings, but thousands. We cannot separate and reflect on all the layers, but subconsciously we understand them. The ability to see cannot be taught, but it is based on the very essence of the neural network: training the ability to see on the best works.
Impressionist Claude Monet painted the Water Lilies cycle of 250 paintings over 30 years. I looked at 240 images of water lilies based on 60 different techniques in 3 hours. A tenth of them made it into the article. Of the 240 pictures, I only got three images that could be used to develop an author's style. They were created by chance, they cannot be reproduced, which means I will train the neural network: create variations of these images, choose and enlarge the best ones.
Over the year and a half of working in Migiorni, I have trained the neural network many times and developed several of my own techniques, many of them complex: for example, you need to transfer an image from one version of Migiorni to another or layer images on top of each other. Here are three prompts that I use in Migiorni versions v6 or v3:
…in the style of Wong Kar-wai
Sam Toft's illustration of a …
Nick Veasey's clean, sketch-style illustration depicting a … –weird 100
I use the resulting image with another prompt in another version to achieve the effect I need. Below I will provide several works in my style.
All the images are made in different techniques and with different means, but they are united by my view: I love subdued light, carelessness and solitude.
If you are interested in continuing to explore the topic of neuroart, I invite you to my TG channel “Parallel editions“.
