Guest
I often see in comments that artists treat those who use neural networks in their art with a touch of contempt. They say that they pressed a button, got the result, and uploaded it to the web. “Here we are, artists, we work our asses off to create a unique work, we studied very hard for this and put a lot of effort into it.“ Here is a natural question: “and neuroartists do not learn anything and do not invest effort?“. Let’s examine this question in more detail. First comes the generation creation. If now the creation of prompt has slightly improved with the appearance of dalle-3 and new versions of Midjourney, then six months ago, to get a decent generation, you had to use dozens of tokens auxiliary tools. And in order to create a scene that we keep in our heads, we have to learn thousands of tokens, keeping in our heads the names of materials, art directions, styles, lighting types, camera angles, artists, special terms, etc. And after we spend half a day generating and finding the image we need, there is a very big step of finalizing the image. We clean up the “errors“ of the neural network, correct poses, finalize the image with HANDS, supplement with details and so on. That is, in fact, we combine the knowledge and skills of artists with the knowledge of linguistic programmers. And to create a decent work, we have to learn not a day, not a month and not a year. For example, to finalize this work, it took me two hours to create a prompt and generate the necessary image and three hours to finalize it. Some works take much more time, especially if there is a complex perspective and pose of the character. So the next time you come to say “neural network drew it“, say “pencil drew it for the artist“ after the comma. Created by prompt from my illustration: the delighted face of a little red-haired boy watching falling stars.
