loader image
What if AI image generators, initially programmed for acceptable tasks, took an unexpected dive into the realm of the unacceptable? What if the cascading effects of AI, designed with specific intentions, started venturing into uncharted ethical territories?

Share This Post

What if we’ve been offering a prism of endless possibilities where a single choice can lead to infinite results?  With open mind I’ve tried several (AI) image generators, watched numerous videos, and read articles on how to write effective prompts. Despite of all this, I’ve struggled to create something that truly aligns with my vision. It seems that whenever someone attempts to use Midjourney or similar models for artwork creation, they encounter a kind of magic trick—a wow-effect with unpredictable results. For me, the challenge lies in the unpredictability.

In my small research, I wanted to try a real life scenario so I decided to replicate a few of my own illustrations that I’ve been paid to create by real client for real books.

Recently, I drew an illustration for Broken Blade Publishing featuring a fantasy character resembling a Landsknecht standing on a ship’s deck with a rifle and a sword, portraying a pirate version of the famous medieval solders.

My illustration for Broken Blade Publishing

From the start, none of the AI image generative models recognized what a Landsknecht is. Apparently, these models haven’t been fed by any images of Landsknechts, despite the great amount of historical reenactment photos and Warhammer Fantasy games where these solders were depicted in great detail.

The result from Midjourney
The result from Leonardo AI
The prompt: A Landsknecht from the XVI century stands on the ship's deck, holding a rifle in his right hand. His left eye is covered with an eye patch, and his head is wrapped in a bandana. Wearing a black jacket adorned with red decorations, he has a long sword attached to his waist on the left side. His right leg is bent, resting on a raised platform, mostly obscured by the ship's deck. In the background, a vibrant blue sky and ship ropes complete the scene behind his back.
Midjourney
AI image generator

In my second attempt, I sought to recreate a cover illustration I did for the same publisher, featuring a giant serpent leaping in front of a castle surrounded by rough seas.

My illustration for Broken Blade Publishing

This time, the Invoke AI (creative engine working on Stable Diffusion models) was unfamiliar with the concept of a giant serpent, and it was very hard to create a dragon without wings and limbs. Even though the result was closer to the desired result. 

The result from Midjourney
The prompt: illustration of a big majestic castle perched on a cliff, with turbulent seas below and a mysterious, dark sky above. Fantasy dragon without wings and limbs gracefully leaping from the sea. Fantasy realm with a smooth and magical artistic style, high detail, fantasy, realistic, light effect, hyper detail, volumetric lighting, cinematic, macro, depth of field, highly detailed, epic, cinematic, concept art, cg.
Invoke AI
AI image generator

It became apparent that these AI image generators struggle to produce creative results. Despite my efforts to distance myself from this thought and give AI a real chance in the concept art and illustration fields, we seem to be stuck where we started

“(AI) Image generators provide service to manipulate someones artwork and often fail to produce original content”
Pavel Guzenko
💡🎮✏️ Artist/Designer

While some models in Invoke offer tools to manipulate details in generated images, their capabilities are not even close to those of real artists. Moreover, they are still limited by what usually described by many YouTubers as ‘AI doesn’t know this,’ which is essentially ‘the model has not been exposed to this image before.’ This raises the question: Is this a creative tool, or is it merely a means to manipulate copyrighted images of other artists to create something vaguely reminiscent of what the user desires?

Giving the nature of fast developing industry I would like to keep this question open.

I acknowledge the vast and diverse field of AI, knowing that some great tools existed even before Midjourney made its debut. Tools like Photoshop’s generative capabilities for filling empty spaces, Nvidia Canvas for producing landscape images, Autodraw by Google for creating symbols from sketches, and even Unreal Engine 5 for crafting beautiful landscapes based on limited inputs from the user. However, these tools lack the magic that the general public expects when hearing about AI.

As of this post’s writing, there is a circulating discussion on social media about some companies that traditionally relied on the work of real artists—such as Wacom and Wizards of the Coast—now incorporating promo art in their ads using AI image generators instead of real artists. This not only downgrades the quality of the artwork but also delivers a significant blow to professional artists whose job has never been very lucrative. Artists certainly didn’t expect the tools they are using, the clients they worked for to turn against them, all for the sake of saving a few bucks.

In conclusion, while many AI tools have the potential to diversify the creative process for artists, it’s crucial to understand that they don’t replace the artist. The trend of using licensed artwork for manipulation and passing it off as original is concerning. Professionals from various corners of the art world are raising their voices against rights infringement, and we should heed their concerns. While it may be easy to get from (AI) image generators a nice image with a simple idea, such as a cute female character, crafting an original artwork demands more than just image manipulations.

you might also like

GRAPHIC DESIGN

Web design, corporate style, motion design

CONCEPT ART

2D and 3D art for video games

ARTWORK

Illustration for books and games