Menu
Tablet menu

Can a Neural Network draw a game for you?

 

Yes, but there are nuances. 

So, Mechahearts is a visual novel in a steampunk atmosphere and it is drawn using the Midjourney neural network with a little bit of Dall-e 2.

I am very skeptical about neural networks: they are still unable to produce a number of interconnected independent works. Neural networks inspire primarily with works scattered over the network in the spirit of this:

Screenshot_Capture_-_2023-08-30_-_22-09-39.png

And objectively, this picture looks very good, but if you look at the command ( prompt ) that the author used, it becomes clear that the neural network did not really adhere to his requirements:

The man specified several times that the girl should be depicted in full growth, that she should have cat ears and yellow hair, but in the end, midjourney gave him an abstract image “on the topic”. And this image is beautiful, but it does not meet the requirements of the "customer".

This behavior of the neural network makes it very difficult to generate many characters in the same style, which is a necessary task for the development of the game. The idea to create a crowd of heroes and stand out in their number against the background of other short stories is repaid immediately: firstly, it is very difficult to get exactly the desired picture through a neural network, even if you always have to rely on luck to pile up three-story requests. In addition, the neural network also rewards the characters with their favorite artifacts.

The eyes are the most affected. You look at this:

Screenshot_Capture_-_2023-08-30_-_22-10-01.png

Everything suited me for this portrait, but I could not get decent eyes. My goal was a nice-looking result, I didn’t want to leave artifacts and glitches that can be forgiven for the neural network, as we forgive them for children’s drawings: they say “so what if this green daub doesn’t look like anything at all? My son painted it, I'll hang it on the refrigerator.

I really liked this picture without jokes, but now I understand that this crooked daub does not carry any artistic value, and it absolutely does not matter who drew it - even a person, even a neural network, even a cat. She looks bad.

I did not want the graphics in the game to have value only for the reason that it was drawn by a neural network, so I began to shamelessly correct all the drawings. I understood that in the end, the user does not care who drew this picture: he cares how it looks.

Screenshot_Capture_-_2023-08-30_-_22-10-13.png

This is an example of the strongest retouching I've done on a character. Most of the other heroes went through less severe edits, I really got carried away here, but I'm happy with the result.

Emotions were another problem - what kind of visual novel is it when the characters show the same facial expression? But any attempts to feed the neural networks of already drawn characters and sign something like “angry” to the promt produced completely absurd results:

This screaming face is very cool, but it does not look like the result that I wanted to get at all. As a result, I again had to try to draw something on top of the result received from the neural network:

Screenshot_Capture_-_2023-08-30_-_22-10-28.png

By the way, I'm not such a great artist, and the style chosen for Mechahearts is too realistic for my skills. With realism, I wanted to enhance the effect of the "uncanny valley" and deprive myself of the opportunity to move away from the intended concept and draw the entire game on my own. I'm used to a more cartoonish style and hard brushes, realism and stylization of "oil paintings" I do not pull. This is how I draw without neural networks This is how I draw without neural networks

During this entire Mechahearts experiment, I came across another almost impossible task: drawing a character from a different angle, and when I needed to portray some of the characters with their backs, I spent a lot of time on a dismoralizing and stupid sorting through requests. I was especially amused when one of the characters began to draw half-ass. I was especially amused when one of the characters began to draw half-ass.

In the end, after completely mindlessly burning through traffic and resources on the Midjourney servers in an attempt to portray my pre-retouched maid from the back, I got an image that I could work with:

Screenshot_Capture_-_2023-08-30_-_22-10-43.png

Neural networks are very inaccurate in details, and as a result, the sleeves, cap and coat hanger disappeared. I had to redraw it all again, or collage from other results:

In the end, it turned out, in my opinion, tolerably, but this drawing is still much worse than the one that a person could create.

But the restrictions imposed by the neural network do not end there either, and sometimes its impenetrable madness forces me to give up and simply not even try to generate the scene that I need in the story. It will be easier for me to rewrite the plot.

I needed a beautiful red-haired girl lying on the floor. Along the way, the idea had to be slightly redone, because neural networks have even worse problems with complex poses than with eyes or hands. A complex pose without a broken anatomy can only be obtained by chance, and it is not a fact that this particular pose will turn out to be necessary.

Working with a neural network is reminiscent of quick blind shooting - it's fun, you are littered with pictures, but 99% of them are rubbish that simply cannot be used in the current project. I used to develop games by myself, not limiting my ideas to other people's capabilities, but now I have to adapt to the peculiar skills of the neural network. I'm adjusting the plot for her. Inventing a story based on possibilities is a normal process - this is how movie scripts are written and games are made, it's just that this approach is unusual for me personally.

Screenshot_Capture_-_2023-08-30_-_22-10-59.png

Does the use of a neural network make life easier for a developer in the end? It all depends on his own wishes and skills. My characters would be creepy cross-eyed freaks with three extra fingers if I didn’t have the skill to edit them, and I have this skill, because I draw for a long time myself, without the help of neural networks.

On the other hand, adjusting a portrait or background is still much faster than painting it from scratch, so I can devote more time to the plot component. Of course, I could choose a style that is more similar to mine and then adjusting the drawings would not be so difficult, but then I would not understand at all why I spend time compiling three-story queries in midjourney, if I myself can do it draw? The trick is just to get a result that is as far from my style as possible.

When working with a neural network, there are no difficulties. When working with her, there are inconveniences.

You can talk a lot about whether neural networks simplify development from a technical point of view or, conversely, only confuse, but you need to put a moral point of view above all, because without the proper fuse and love for your project, no matter how technically cool and advanced it is would not be, no game will be completed.

Development should be fun for the brain to perceive the work on the project as a reward. Otherwise, he will always try to force you to do something else, and creating games will become painful work.

When I paint, even if I do it on a table, I feel like I'm growing as an artist. Without those hundreds of drawings that I crumpled and threw away, I would never have gotten the image I wanted. And that sense of development is where the fun lies: my brain knows that I have to change and improve in order to get what I want, and it rewards me with dopamine when I do.

Screenshot_Capture_-_2023-08-30_-_22-11-17.png

In the case of working with a neural network, I do not develop at all. The wording “prompt-artist” is, IMHO, nonsense, which should just create the illusion that you are gaining skills and developing as a specialist while sorting through various requests or even repeating the same one 50 times.

But it is the neural network that is learning and developing, not the promt-artist.

The selection of a query is a very boring and illogical puzzle, the solution of which does not make you better or smarter, but solves a one-time task. The experience of this decision is hardly useful next time. Generating pictures is easy physically, but mentally exhausting, because the feeling that I'm wasting my life and not gaining skills is very vile. Probably, for this reason, I correct these pictures with such pleasure, because that's where creativity begins. And on this good-natured note, I finish the story of how a novel is made using a neural network. What the game itself is about and what distinguishes it from other short stories - I will write next time.

Last modified on
>
back to top