Neural Network Art: How an Artists Can Use It

If you are impressed with neural networks and their capabilities, then you will probably be wondering what else they can do and whether they can ever replace a human.

Does AI have a soul? Find out more here.

In this article, we’ve tried to consider the topic of neural network art. Read about this and many other things here.

Have you ever imagined what a digital person would look like? Read our thoughts here.

How neural networks work

A neural network is a concept from biology. The neural network device is very similar to the brain: there are neurons and connections between them. Neurons receive, process, and output information, and connections transmit it. 

Neural network working process

The difference between a human brain and a computer neural network is that neurons in the brain can independently form stable connections when a person learns something new. And neurons in a computer network do not know how to learn independently and form connections. They only know how to receive, process, and issue information on command.

For the neural network to start working, it needs to be taught. It’s like explaining new things to young children. Here you show the child pictures of animals and call them: cow, cat, dog, and camel. Then the child, when asked to “point to the camel,, shows you the correct picture.

In Chile, a law on neurotechnology was approved. Find out more here.

The neural network also works the same way. When processing information inside it, a complex process takes place. But the principle of training and subsequent execution of the command is the same.

The biggest misconception is that neural networks can learn independently. Remember Ultron in the Marvel Avengers movie? He began as an artificial intelligence that gained access to information, went online, and became the world’s strongest intelligence. And then he constructed a physical body for himself. So, Ultron’s self-study is an artistic fiction. All modern neural networks are trained by humans. And the algorithm of their work is written by them.

Each neural network has its task and its algorithm. For example, Google image search is also a neural network. You show the picture and click the “find” button. For a neural network, this is a command — “find me the same one or anything that looks like it.” A database of images is preloaded into Google’s memory, the network takes yours and runs it through its database, comparing it. As a result, it gives you everything that it considers similar. How to run through the database, and which pictures are considered similar, the neural network was taught in advance by a person.

But the DALL-E network independently generates an image from elements known to it according to a text description. You write “an astronaut rides a horse, a pencil drawing.” The network itself searches for suitable images, combines them, adds a given image style and produces a ready-made picture.

We wrote about DALL-E capabilities here.

Neural networks are still very imperfect: sometimes when creating an image, something meaningless or even frightening turns out. Or when searching for a picture, what the network gives you as “similar” is actually not at all similar to the source. This is because the neural network is so trained. And for artists, this means that a neural network cannot yet replace a person.

Neural network work

Neural network art tools

#1 Disco Diffusion

Neural network from Google. It creates pictures based on a text description: you need to come up with a theme or plot and send it to the network. In response, it will draw a picture based on its knowledge of the given plot. If you repeat the same plot again, the picture will be different.

How to use: there are many settings in the program, but not all of them are needed. First of all, find item 3 — settings:

Neural network tool setting

Select the number of steps in the steps column. This is how many images will be used for the final result. The more steps, the sharper the image. But we do not recommend choosing more than 500; otherwise, the picture will be too detailed and will take a very long time to be generated. Usually, 250-300 steps are quite enough. In the “width_height” column, you can select the size and format of the image.

Next, in the “Prompts” section, enter your query about what the neural network should draw. In addition to the topic itself, you can enter either the name of the artist/artists whose style the neural network will rely on or trending on Artstation — then it will rely on the works from the “Popular” page in Artstation:

Neural network tool

After that, go back to the menu at the very top of the page and follow the path: Runtime –> Run All. Or press Ctrl+F9.

The network will ask you for access to the disk and get to work — usually, it takes from 1 to 3 hours for one drawing.

#2 Artbreeder

Mixes different images together. You can take several portraits and get a new face. And you can also change gender, age, mood, facial features. However, you can take portraits only from the database — you will not be able to upload your own images.

How to use: To generate a new picture, click on the green plus sign in the upper right corner of the screen. Then in the window that opens, select the theme of the future picture.

Then click on the “Add pattern” button and select the images to be processed. Play with the settings to get an interesting image.

#3 Nvidia GauGAN

This is, in fact, not one neural network, but two. One of them creates images very similar to those in the neural network knowledge base. The second determines whether it is unique. The two neural networks constantly seem to compete with each other. Therefore, this neural network is called generative-adversarial.

How to use: Select a picture theme and draw on the left side of the screen.

When the drawing is ready, check the boxes at the bottom of the workspace to agree to the terms of use. Then select the image style and click on the right arrow button. On the right side of the screen, you will get the result.

#4 EbSynth

A network for styling and changing images or videos. Let’s say you want to change Arnold Schwarzenegger’s uniform in the movie “Terminator.” To do this, you need to take one frame, save it, redraw it the way you want, and run a command to process the video according to this frame.

How to use: The process is quite time-consuming and requires preparation. You need a video and a storyboard in the PNG format. Select one frame, process it and save it to a folder. The program synchronizes the processed frame with the video.

#5 Dream by WOMBO

This network also generates images based on a text description according to the specified templates. The result is abstract drawings.

How to use: In the Enter prompt column, write a query, select a style below, and click the Create button.

How artists can use neural networks in their work

Neural networks are tools that will be useful to an artist.

  • You can use the image created by the neural network as an underdraw and draw on top of it.
  • You can arrange a battle between an artist and a neural network, when an AI and a human draw a picture on the same topic. Then, by voting, choose who handled the task better.


Drawing using neural networks can be quite fascinating and allows even amateurs to create pictures. Neural networks are another new tool that people should not be afraid of.

But even if you use neural networks, you need to pump up your basic skills, work with color and light, and be creative. Without these skills, drawing even with the help of a neural network does not guarantee a good result.

Humans will always be one step ahead of neural networks because it is humans who can come up with truly unique ways to present familiar topics.


Leave a Reply

Leave a comment

Your email address will not be published.