Embedded Files
I don't intend for this article to be an introduction into what AI is nor the moral/practical qualms around it in terms of stealing designs from artists without consent, nor the environmental impact of it. These are very real factors, but others with more relevant experience have already discussed this at length in other places. But if you understand these factors at length and have been educated on them, it's unlikely you'll be at this page to begin with, as you probably wouldn't be considering using an AI reference.
Tattooing is my domain of expertise, so in this article, I'll be focusing on why it has never been useful in visualizing nor discussing tattoo projects. Again, the other reasons to avoid it are valid, but environmentalism is not my expertise, and the rest is what I want to focus on for this article. I'll be using some examples of past client interactions, and discussing the unnecessary difficulties we faced because we spent so much time and energy breaking down the errors of AI references.
Example 01: Bug Rework
(Note: I will be changing some smaller details about this story so that I'm not pointing out any project in particular.)
Let's call this client Ava. Ava approached me with a rework project. Ava had a bug on her wrist that she felt was not tattooed very well (blown out lines, fading color). She wanted to rework the bug in a more realistic way, and add some flowers and a dagger above that bug.
Ava sent me a ChatGPT image for visualization. I was told it was just a loose and rough reference, and that I was being trusted with any necessary changes.
I noticed it made a large number of choices that would just unfortunately not be possible to translate into Ava's tattoo project. I spent almost a dozen paragraphs explaining what wasn't possible, and offered a new sketch as an example. I explained what worked well about the new project, including finding an S-shaped gestural flow (indicated in green), considering the flow of her body and the old tattoo. But we still faced a lot of obstacles in undoing all of the AI's mistakes. Let's explore why.
What issues did the AI cause in this example?
It didn't take limb proportions into account.
When I design tattoos, I am incredibly intentional about finding a sense of flow with the body. I also frequently use client photos (especially if we're working around other existing tattoos) to make sure that the proportions of the tattoo align with that person's limb or body area.
Limb proportions are not just about width and height. For some people, their limbs taper more when going towards a wrist, for example. Some people have thicker thighs, thinner calves. Some have thicker calves, thinner hips. Some people's arm will be very round at the edges, or very straight. Some people have a lot of muscle and a lot more 3D shapes to account for, and more visible "bumps" of the muscle on their limb's silhouette. These are all things I take into account when designing for my client. Some people have scars to work around, other tattoos. All of this is vital to how the final product will look on the body.
The AI completely ignored what my client's proportions were. It made the arm unnaturally long and thin compared to my client's. My client had a more pronounced taper from the top of the forearm, a bump from that brachioradialis muscle that was visible. That was important, because when designing the proposed project, I could create some visual interest with a little bit of size variation between the flowers. If I had simply used the AI's proportions, it would have left huge gaps at the top of the arm.
2. It didn't take the arm bend or areas to avoid detail into account.
As tattoo artists, we are trained to know which areas are best for focusing detail, and in which areas detail should be avoided. For example, on the elbow bone and right in the elbow ditch, we get a lot of distortion - because it's right on the joint, any movement or bend will distort things within it. So an object like a dagger that relies on looking as straight as possible to look good, it wouldn't have been a good idea to place the dagger right on the bend. The AI didn't seem to give any indication that it even understood that there was a bend at all, much less having the understanding to plan around it.
My client had also asked to not go above the forearm, so at the proportions that the AI was offering us, it was just not possible to fit that only in the lower arm while also having a similar distance to the sides of the arm. It would also have not been possible to do so while keeping similar proportions to the existing bug tattoo.
I sent a similar visual to my client, but putting elements like daggers on that bend is generally not a good idea. When doing sleeves, artists tend to break up the detail in that area, potentially with wind bars, or other elements that can be simpler.
3. It didn't create a good gestural flow with the body.
Bodies are not symmetrical. Even when there is an impression of symmetry, there rarely is true symmetry. And while symmetrical designs can look good, designs with a gestural curve that can follow the flow of the body's muscles (either with a C-curve or an S-curve following the elements in a general sense) almost always look better.
We mentioned the brachioradialis muscle. This means that forearms usually have one side where the peak of that curve is a bit higher on one side than the other. And because of how much forearms rotate, there isn't usually one true "center" to it that remains a clear center during movement. With c-curves and s-curves, there's usually an organic sense of flow where elements still make sense even during that movement. But more extremely symmetrical designs can often suffer from movement.
You may have seen a trend on social media with "twist tattoos". it isn't anything inherent to the tattoo necessarily, but simply the recognition that tattoos that have a straight gestural line (like daggers, or long lines of text) will twist with the body. Forearms simply happen to be an area where we see that more extremely compared to other areas, as it has a greater range of motion.
On top of the consideration that we are not working on a static canvas, but skin that moves, a symmetrical centered design didn't make sense as the bug tattoo was in a 3/4 pose. Which brings us to our next point:
Kenzie Carter on TikTok.
Tattoo by Dominic Dafoe.
4. It shrunk down and re-posed of the tattoo that was already there, which is not possible with a rework.
Generally speaking, whenever covering up older tattoos, dark colors can only be covered with darker colors. This was a tattoo with thick and black outlines that was already at the edge of the wrist (not much space for a different design), which is why our most optimistic outcome was doing a rework where we honored where the old tattoo was and simply added more nuanced shading and details. There can be some minor exceptions when a tattoo is 10+ years old (like covering black outlines with a dark purple or a dark blue), but shrinking down and moving the tattoo would basically require removing ink. That is not possible at a tattoo shop - that would have only been possible with extensive laser treatments, which the client was not open to.
The client was very attached to the symmetrical design. However, our bug was much larger and in a 3/4 pose. I informed the client that, especially for a tattoo under 10 years old, simply overlaying the new bug pose would read more as a blastover tattoo (example on the left). In blastover tattoos, artists intentionally overlap ink knowing that the old design will show through. But I knew this wasn't my client's intention. And so they were informed that either we had to honor the original pose and proportion (working it in a much more satisfying way into a gestural curve as it wasn't symmetrical), or we had to commit to laser removal if we truly wanted the smaller bug in a different pose.
Which brings us to what was ultimately our biggest issues...
5. It created an emotional attachment to an unrealistic version of the tattoo project.
I was told from the beginning that the client had full trust in me to edit the AI image. I was told that they understood that it could likely not be tattooed as is, and that they accepted my expertise and input. However, every step of the way, there was significant pushback.
Can we not truly make it symmetrical? Can we not truly keep the dagger vertical without shrinking it? Unfortunately, this was not possible, as she simply didn't have much space left over in the forearm.
Many people who have worked with me know that I try my best to offer options. More often than not, I send people at least 2-3 thumbnails of poses and compositions to choose from before we proceed. I give people color options whenever applicable. I send designs ahead of time for significant review, which is not standard in the tattooing industry. But I was struggling to offer options because my client was attached to only one - the AI version that they had already pictured in their head. Even when being given revision round after revision round with 2-3 options each time for every element, nothing was good enough because it wasn't what she was already imagining long before she contacted me.
Even with the willingness in theory to give trust and cooperate, we had to spend days and paragraphs upon paragraphs breaking down why the AI image couldn't even be a starting point before we did any work on the true design. And while Ava was very content with the final work, we wasted so long just discussing point by point what AI didn't understand. If we had worked with each other from the ground up, there's a good chance our process would have been smoother and involved a lot less disappointment if we had simply used reference images for aesthetics and vibes and discussed what the goals were together.
Mo Eid on Pexels.
Pavel Danilyuk on Pexels.
6. It discourages active participation in thinking about the parameters of your project.
When using an AI, there's an appeal to "just seeing how it looks" and feeling it out based on what you see. While I do think visualization of the project is vital (which is why I do like to send sketches and designs ahead of time), there is a very beginning stage of figuring out what you want while actively thinking about it.
For this point, I'll be using a different example with a much simpler case.
Example 02: Cupcake
(Note: I will be changing some smaller details about this story so that I'm not pointing out any project in particular.)
Let's call this client Beatrice. Beatrice approached me with a simple project, a cupcake matching with a friend. I sent over some reference images as this subject matter has been tackled plenty in the past, and asked questioned about the style. Beatrice wanted it in color, and said she wanted it as a character.
Before I could ask further questions about how she wanted the personification executed, I was sent an AI image that she said her friend made. She said she liked it as is.
In the AI image, the cupcake was done in a glitter style tattoo (example on the right for those unfamiliar). It was also done in pastel colors, and as my client had a darker skin tone, I explained that I would have to deepen the colors. She understood. I double checked to confirm that she wanted the glitter style, and I was told no, not at all.
While I do always double and triple check on parameters as much as I can, If I had not asked, we could have had a disastrous situation. She could have gotten a style completely different from what she wanted because she had not paid attention to the details of the AI image. Its appeal of offering a simple solution where you don't have to workshop it, where you can just have something be given to you, has a massive downfall - it means that parameters of the project (size, style, lighting, line thickness, placement) aren't fully thought through. And when it comes to permanent body modifications, unfortunately, there's no backing out! Once ink is down, it is down. And if something was not fully thought through because there wasn't a willingness to think everything through, we risk getting a result that could be regretted forever.
Additionally, this was just a cupcake with a smiley face on it! If you don't trust your artist to draw something that simple, then they shouldn't be trusted to execute a permanent body modification on you. I could have been given the same "prompt", and come up with something that was even more accurate image to what my client wanted and what would look good as a tattoo. Remember that you are paying your artist to design your tattoo - this is included in the final price you pay. So take advantage of the fact that you have an artist at your disposal with the training to make designs tattooable!
An example of a glitter style tattoo by Amanda Graves.
So what's the solution?
The solution is tried and true -reference images + descriptions!
Not every reference image needs to be a 100% identical match to what you are picturing. But there is where words come in too as a tool on top of images. Take 3-5 reference images, and describe what you do and don't like. Think of parameters like:
Do you like the placement?
Do you like the style?
Do you like the line thicknesses?
Do you like the subject matter?
Do you like the secondary and tertiary elements (things that surround the main subject)?
If there's flora or fauna, are there specific species you want (like a specific flower, or butterfly type)?
Do you like the silhouette?
Do you like the color scheme?
Do you like how light/dark the tattoo is overall?
For example, let's say a client named Callie wants a neotraditional chest tattoo of flowers. She sends these images and descriptions:
I love these colors and the leaves. I don't want the ornamental elements in the middle. But I do want the flowers to be peonies or similar to peonies.
I like that the flowers are coming from the outside in, and I like that they go over the shoulders. I don't want it to extend to the neck though.
I don't like the direction of the flowers, but I love the thickness of the lines and the flow of the petals.
I really like the addition of the figs, although I don't find them to be a necessity.
This is beautiful because it is:
specific where things are specific in her mind ("peonies" vs just "flowers", the color scheme), and honest about what elements are more flexible (like the figs not being a necessity)
it focuses on visual elements (e.g.: thickness of the lines, not just vague terms like "pretty" or "elegant" which can mean many different things to different people)
has boundaries when relevant (like not going to the neck but okay going over the shoulders, or not doing ornamental elements in the center)
Your descriptions don't need to be perfect or overly technical by any means. Our interaction at the end is a conversation, and I'll be able to ask clarifying questions if I feel I need them. However, what is important is taking time to really think through what you want! Don't rob yourself of the ability to connect with your own tastes and desires for your presentation and your body, as no AI can give that to you.
Google Sites
Report abuse