All artists sooner or later begin to take orders for drawing illustrations. Customers send you technical specifications, which are a description of what you will need to draw. Your task is to figure out what the person wants. Not all customers give a detailed description; sometimes the task is extremely incomprehensible, and many hours of correspondence and calls do not help.
What should you do to understand what the customer wants to see in the illustration? Yes, it’s easy, you can turn to neural networks!
Now there are neural networks that can be used to create illustrations based on words. On these networks you can simply write the keywords from the technical specifications and get a picture. It is important that this will not be your finished work; from the resulting picture you make a sketch, or you can take a composition or a color palette. After processing it, you can come up with options for the customer.
If you now shout “what are you talking about, you can’t do that, it’s cheating!” , no, friends, this is our reality, now everyone loves speed, the faster you do what the customer wants, the better, and if you use a neural network as an auxiliary tool, on the contrary, it will make you a more modern artist.
I gave you a test drive of neural networks.
I have chosen the simplest and free neural networks for you, I don’t like to spend a lot of time understanding the settings, well, I’m not a techie, but a humanist, I need, first of all, nativeness so that I can immediately understand where to click.
I chose three neural networks, because you can get confused in them, there are too many of them.
Let’s imagine that the customer wrote you a technical specification (specification specification).
“Good afternoon, my name is Maria, I want to order a beautiful illustration from you for my travel blog, I want a portrait of myself against the backdrop of the mountains with a cup of coffee in my hands, I also want flowers in my hands or nearby. I want the colors of the painting to be rich, bright, spring-like, to set the mood for positivity.”
Here’s a task, now I’ll pass this description through neural networks, and after that we’ll see what I can come up with? Will I get the result I wanted?
Selects keywords from the task description.
Words in English: love, journey, mountains, coffee, girl, flowers, painting.
1. Neural network Stable Diffusion
The first neural network “Stable Diffusion” (Stabil’naya diffuziya stable difusion) – it converts text into images, capable of generating photorealistic pictures. When you enter text, 4 photos are generated, so you can make a unique reference for illustration.
Result: An excellent result, this is exactly what I wanted to get from the network, something I can work with, take ideas and redo, but the network did not produce good results the first time.
2. Kandinsky neural network 2.0
The Kandinsky 2.0 network is a model that generates colorful images in a matter of minutes. It can use different directions in painting. The network from Sberbank also works through the Salute application, through the “Turn on the artist” command.
Here’s what happened:
Result: The first option is in the style of Aivazovsky, the second option is in the Renaissance style. Interesting results, the network is still based on the paintings of artists, well, it cannot be created out of nothing, it still needs a basis, a foundation, without this you can’t go anywhere. They turned out to be good works, but unfortunately they are not suitable for my task, they are too similar to oil paintings. Since they also have rights from Sberbank, they need to be thoroughly redone so as not to get into trouble
3. Craiyon neural network
The third neural network, Craiyon, is a neural network that generates content based on a text request in English. The interface is easy. I inserted the words and the program generates 9 illustrations at once, it’s cool that you don’t need to click several times.
I tried this network, here are the results)
Result: The illustrations turned out strange, and I added the words picture, but not for children, and even after several times the result did not change. The cup is generally separate in several paintings. Perhaps this is still an unfinished neural network.
Let me summarize my test drive:
I liked the results, there is plenty to choose from. If the task were not an example, then neural networks definitely helped me, and most importantly, saved me time, which is very important for everyone.
Go ahead, trying neural networks is very exciting, the main thing is not to get too carried away and waste time. You need to go to neural network sites with a specific request, know what you need, what you want to get as a result, so as not to go into the wilds.
Bookmark it so you don’t lose these neural networks! They will definitely come in handy! After all, while you sleep, they learn to be better!
