Artificial intelligence that can transform anything that comes to our mind into pictures in the style we want; It reveals very interesting results when the subject is hand and foot drawing. Sometimes it adds extra fingers to people, sometimes it leaves them missing, and even the fingers are not where they should be.
If you are like me; If you are an expert and trained in visual arts, you may be guessing the reasons for this failure of artificial intelligence. However, in our world where technology has developed so much, this is a surprising situation for all of us.
Even to the point of dropping artificial intelligence into the language of social media…
“And if you never learn to draw hands, AI can’t copy you and make hands right.”
“Expectation: AI art can draw anything!
Fact: The art of artificial intelligence in hand drawing.”
“Being bad at hand drawing: Me and artificial intelligence”
Let’s take a look at the hands and feet drawn by artificial intelligence programs. Is it really as bad as it is exaggerated?
Hands drawn by Leonardo.Ai:
Feet illustrated by Leonardo.Ai:
Hands drawn by Dall-E:
AI, we get you, because you’re not the only one struggling with drawing hands and feet.
If you have tried to draw a person before, you probably had a hard time drawing hands and feet. Because even when you go to a painting course (if you don’t have any training in this subject before), you are asked to draw hands and feet only as a silhouette in the first place. In order to detail the drawing and put it all on paper correctly, first of all, it is necessary to understand the human anatomy, to know angles and dimensions, and to have knowledge of proportion.
In this way, you can understand from which muscle or bone the indentations and protrusions you see on the model’s body originate and draw a more accurate picture. But when it comes to drawing hands and feet, it’s not just artificial intelligence; we humans can have a hard time too.
The reason of this; The hands and feet have movable and highly detailed parts (like the fingers), although they cover a very small part of the human body.
It is not easy to draw these limbs, which occupy a very small area of the human body and have a lot of detail. Because you need to know the skeletal and muscular systems of the hands and feet separately and to be proportional to the human body. The fact that each person’s hand and foot sizes and finger lengths differ from each other is another confusing issue. As such, the more fingers there are, the more confused the artificial intelligence programs that transform the text we write into a picture within minutes.
Of course, artificial intelligence cannot accurately complete or imagine objects like the human brain. This causes missing or extra fingers or not being in the right place in images with more than one hand, even if it has learned beforehand that “A hand has 5 fingers”.
The human brain, on the other hand, can correctly complete the objects it sees. For example, you may not have noticed the extra fingers when you first looked at the image (above) that we had Midjourney draw of the Uruk-hai race in the Lord of the Rings universe. The reason why we often do not notice the mistakes in the visuals created by artificial intelligence is this little game our brain plays on us. Just like we realize as we get older that the characters in the cartoons we watched when we were little, had 3 or 4 fingers.
In portraiture, it is not as unsuccessful as hand and foot drawings.
You can draw a skull and put thousands of different portraits on it because although the dimensions of our skulls change, the shape does not change much, except for exceptions. The eye sockets, nasal protrusion and mouth are usually always in the same place.
Moreover, the skeletal system in our heads is less mobile and less detailed than our fingers. So where our eyes, nose or mouth should be is much more predictable.
Let’s take a look at hand and foot anatomy.
We have 3 different knuckles for each finger on our hands (except for the thumbs). In order to draw a hand, every curve, indentation and protrusion must be placed anatomically correctly.
In other words, these countless details in our hands and feet, our musculoskeletal system, seem quite complex to artificial intelligence. Moreover, as we have just mentioned, we all have different hands and feet.
For this reason, it is necessary to add a little creativity to drawing hands and feet.
Artificial intelligence programs generally work on learning. It can process and analyze data, but cannot produce successful results when creativity is involved.
Asking the AI to draw only one hand is more likely to succeed, but it cannot creatively complement the size and posture of all fingers on the hand in more complex patterns, such as two people holding hands. So AI can give you something that already exists, but don’t ask it to be creative.
Another AI program asked ChatGPT, “Why are AI programs so bad at drawing hands and feet?” We asked him, and he explained why they couldn’t draw hands like an artist as follows:
So, can artificial intelligence be successful in drawing hands and feet one day?
Midjourney A National Football Team drawing.
It is not very difficult to draw a hand alone, but it is quite difficult to fit it into a whole and convey the details correctly. How many hands, feet and fingers and fingers are in the text you want the artificial intelligence to draw; his head will be just as confused. For this reason, we can help them make a successful drawing by giving more detailed information about the hands and feet in the prompts we enter.
With the development of artificial intelligence technology, it is of course possible for them to produce flawless drawings one day. However, it is not possible for him to be able to make creative drawings with an aesthetic understanding. Because it should not be forgotten that they are only programs related to learning. We already know that even though they’re programmed, they don’t always act “right”.
478
