Why Can’t AI Draw Hands?

Written by Adam Morris

Updated August 6, 2023
Hand on top of paper with sketch drawn

In recent years, there has been remarkable progress in Artificial Intelligence (AI) which has resulted in its ability to produce remarkable pieces of art, music, and literature.

However, there is one area that AI seems to struggle with – drawing hands. Why does it seem difficult for advanced technology to replicate one of the most important parts of the human body, despite being able to mimic human intelligence accurately?

Drawing hands is a challenging skill that has been practiced by artists for centuries. Despite their experience, even skilled artists have deemed it to be one of the most difficult things to draw. It’s no wonder that AI struggles with this task. But why is it so hard? And why can’t AI do it, despite its advanced algorithms and processing power?

This article will discuss why AI struggles with drawing hands and how it affects artistic creation in the future. We’ll examine the current limitations of AI technology and explore ongoing research to address this issue. Let’s begin our exploration of why AI has difficulty drawing hands.

Six Fingers of AI

The recent phenomenon of AI-generated artwork has been truly remarkable. AI-based tools like Midjourney and DALL-E have enabled anyone to quickly and easily create incredible AI art with a few simple clicks.

Group of hands huddles together with one hand on top having six fingers

Online communities of over 100,000 people have grown rapidly around these tools, sharing their creations, discussing the technology behind them, and more. Even leading art community ArtStation was forced to address artificial intelligence’s influence on their site, prompting an artist strike that sought to label art made by algorithms apart from work made by hand.

An interesting and unsettling quirk of these AI tools is the tendency they have to produce images featuring hands with six fingers or more. While it’s easy to write this off as a curious anomaly, it’s actually indicative of a larger issue: why can’t machines get the simplest human representation right?

This is a problem faced by those creating AI models, especially when dealing with concepts related to biology – but there may be no answer to this mystery until researchers find more accurate ways for machine learning algorithms understand extremely complex biological structures like our hands.

AI Image Generators

Midjourney is an independent research lab that is working on developing a specialized AI program, aptly named Midjourney. This tool provides users the ability to create images from text descriptions, with the help of a chatbot on Discord. Currently, the tool is still in its open beta testing phase as it was only just released on July 12th of 2022.

Within the industry, there are several analogs to Midjourney like DALL-E and Stable Diffusion. All tools employ similar concepts, which focus on the use of AI in developing image recognition software with minimal human intervention–the difference being only in style and level of development within each one.

Overall, these tools provide exciting functions that assist both researchers and everyday people alike with image recognition tasks they may encounter.

Issue On AI Hands

AI has long been acknowledged to be capable of performing various tasks that were once considered impossible for a digital entity but the one area where AI struggles is in accurately recreating human fingers or toes. This lack of detail or accuracy has become the source of much debate and even ridicule amongst researchers and consumers alike.

So why are computers so unable to recreate something as seemingly simple as our digits? As it turns out, there’s already an answer – and it also requires another form of AI.

Recent advances in generative adversarial networks (GANs) have allowed developers to create intricate images depicting each and every finger – down to the minutest details such as fingerprints, cuticles, knuckles, wrinkles and joints.

Through training these networks with 3D models available online, computer programmers have employed GANs to finally generate realistic images of fingers and toes. So while AI isn’t quite there yet in terms of perfecting the portrayal of human digits, recent advancements have paved the way for increased accuracy that could take us many steps closer towards a flawless recreation.

The hand is a complex part of the body

The hand is a complex part of the body and can be incredibly hard to recognize, sketch or animate. The anatomy ranges from fingertips that come in different shapes and sizes to wrinkles, folds and shading on the palms. Despite this complexity, hands are an essential part of conveying emotion within artwork and must be represented accurately to ensure a realistic outcome.

Female manicured hands on beige background

Artificial intelligence (AI) has had difficulty interpreting hands accurately due to the intricate geometry found in them, meaning there are no existing standard sets of lines or shapes that it can pick up on to properly identify a hand’s shape. Instead, AI needs to manually compare among many different shapes until it is able to successfully produce a realistic looking six fingers for example.

Consequently, this requires much more time and effort than modeling with other parts of the body. Nevertheless, representation of hands play an important role in artwork today and artists need to have a solid understanding of fractals as well as human anatomy in order to get it right every time.

Hands Are Naturally Difficult To Draw

Skilled human artists have recognized drawing hands as one of the most difficult things to draw. Depicting hands requires a sense of precision and accuracy which can prove difficult to achieve.

Not only is their geometry difficult enough to illustrate, but also the subtle nuances in shaping fingers and nails that make them look realistic can be challenging too.

Drawings of hands in different positions

This isn’t something just related to modern technology or advances in artificial intelligence as we often see cartoon characters being depicted with just three fingers and a thumb in order to simplify the task and make it more achievable.

While this may not be particularly realistic and obvious, it works perfectly fine as an artistic solution and has become so widely used that most of us don’t even realize when characters have this simplified design.

Chat GPT’s Opinion

ChatGPT is an artificial intelligence chatbot launched by OpenAI in November 2022 that has been designed to answer queries on a variety of topics. It’s become known for offering detailed and precise responses that have impressed many, even leading some programmers to joke that this bot may eventually replace them one day.

Many people have reacted positively to ChatGPT, citing its accuracy and the depth of its knowledge as reasons why they might consider using it.

For example, if you ask ChatGPT why people find drawing hands and fingers particularly difficult, it provides an accurate response saying that the intricate details and complexity make it challenging for an artist. It goes on to explain how hands are composed of small bones, muscles, and tendons which require considerable skill to portray realistically.

Due to its capabilities coupled with its natural language understanding AI technology, ChatGPT has become quite popular among everyday users seeking answers from various fields of knowledge. Its ability to instinctively provide informed advice has earned it a status as a trusted source of information for many people seeking guidance.

With its growing user base, continuing development from OpenAI, and additional data sets expected over the years; we at least can expect good things from ChatGPT for the future.

How to make AI draw realistic hands better?

Provide activities for the hands

Hands are one of the most complex body parts, yet they have been successfully used to train AI algorithms. This is because hands often take on a variety of shapes when performing tasks.

For instance, if an AI algorithm is trained with the task of holding a cup, it will be more accurate at recognizing hands in this position than if the hands had no assigned purpose. With training data narrowed down to specific hand positions necessary for particular tasks, AI algorithms can recognize these poses much faster and more accurately.

Barista holding cup of coffee in her hand

However, not all AI generated images of hands are successful. We can see this in examples like DALL-E and Midjourney which both generate images featuring a person holding a glass or mug.

While the photo of the girl holding the glass seems realistic enough based on what we’re accustomed to seeing when people hold something in their hands, there are still bumps and irregularities visible that make us question its accuracy. Still, despite such imperfections, AI-generated images continue to prove useful within certain scenarios.


Inpainting is a useful strategy for image editing, and has particularly valuable application in redrawing parts of images. By using Inpainting techniques, a computer can erase part of an image to fill the gap with something else. This can be used to neaten up compositions or remove imperfections from an image.

One example application is redrawing hands in an image. With most traditional retouching techniques it can be difficult to get detailed and convincing results when working close to other elements – such as faces and fabrics – but with Inpainting they become much easier.

The Dall-E 2 program has been designed especially for this task and can produce great results even in complex images. It will work out which parts of the picture are connected and decide how best to fill in the gaps, creating realistic results that look like the hand was never removed at all. Comparing before-and-after versions of inpainted photos shows the impressive transformations that the Dall-E 2 program is capable of producing.

Use Other Editing Programs

One way to improve your digital artwork is by taking advantage of an AI generated hands. While these hands may already look nice, they can still be improved with a little bit of help from a graphic editing program like Adobe Photoshop.

With some simple techniques such as selecting and unselecting parts of the hand, manual blending or adding colors and textures, you can considerably enhance the realism of the finished product.

Male sitting in front of computer with Adobe Photoshop on screen

You may not always need to rely on software to improve your artwork though. If you have some artistic skill, you can also take matters in your own hands and do the improvements manually. With the help of tools such as markers, erasers, pencils and other drawing supplies, you can make sufficient changes to develop more detailed artwork without having expert knowledge in editing software or AI-generated images.

This is especially useful if you feel limited by any potential legal issues when using certain copyrighted or open source graphics or templates. Whether it’s through manual work or via software manipulation – it’s solely up to you which solution works best for your project!

Cropping Image

Cropping a photo is often the most sensible solution to making it look better and more aesthetically pleasing. In this case, one user on the Midjourney Discord server decided that was the best approach for their photo. This user found that by simply cropping out some of the hands in the frame, they were able to improve its overall appearance.

The difference between before and after was remarkable. The simple process of cropping improved the photo in ways that drastically increased its quality and made it more suitable for whatever purpose it was intended for.

By taking just a few minutes to crop out the unnecessary parts, this user saved themselves hours of editing time, as other methods may have only been partially effective or not effective at all. It’s clear that sometimes, even such a simple adjustment can make all the difference when it comes to photos and visual media.

Use Image Comparison

Image-to-image comparison is a great feature provided by Midjourney that helps AI interpret images. By providing them with a photo, the neural network can process its content and better understand what it is being asked to do. This method makes it much easier for AI to create hands as creating them from text alone can be difficult.

Female artists painting seashell with reference image on phone

A common problem when creating hands is accurately interpreting what has been written; providing 5 more hints makes it significantly easier for the AI to arrive at the correct result.

For instance, simply writing “hand” might not give us the right outcome, so adding more details such as shape and size of the hand, color of skin or nails, or even jewelry worn on the hand will give AI more information about what should be created. Adding these additional hints drastically improves the accuracy of the creation process and allows users to get exactly what they want from their project.

How else can we make AI draw a hand?

AI technology has made it possible for us to create interesting and realistic images with less manual effort. We can now use AI tools to draw a hand in a much faster way without sacrificing quality.

One of the ways we can do this is by training an AI algorithm with various images of hands, especially mittens, which are easier to recognize and draw because they have five fingers and are symmetrical. This helps the AI learn how hands should look like, as well as the right proportions, shadows and colors.

The same concept can be applied when drawing hands involved in an image or action. For example, if you need to draw a couple holding hands while walking in a park and wearing gloves, all you need to do is provide the AI with the keywords “couple holding hands walking in a park and wearing gloves.”

With these easy-to-understand words, the AI can begin building an image of what you had in mind. Of course, if you zoom in on the image closely, some details may not look as perfect as you would hope them to be but from afar, most people wouldn’t even notice any small imperfections that might exist.


In short, although we have seen some instances where artificial intelligence can depict a hand with 5 fingers, this isn’t always going to be accurate. This is because the structure of a human hand is very complex, making it difficult for machines to get right. In addition, due to its relative newness, even experienced artists may struggle to quickly create a realistic depiction of a human hand with AI-assisted tools.

Nonetheless, there are enough options now available so that machine-learning models can accurately depict hands with five fingers and other complex structures as requested by people. These neural networks continue to rapidly evolve in capabilities as well. As such, it appears that machine uprising will not be an imminent reality anytime soon.

Adam is a crypto expert & AI enthusiast who has been researching and writing on the topics since 2017.

He’s spoken on numerous podcasts and has been featured in many prominent media publications such as Forbes, CNN & CNBC.