My stance on A.I.
I co-founded a video game studio and spent several years working on Little Legend, a project driven by our ambition to create the most beautiful 2D video game ever, emphasizing beauty and aesthetics to touch the very depths of the human soul through beautifully hand painted visuals…
…and then A.I. entered the scene.
Before getting technical, allow me to share my story with you.
I come from a family of artists; both of my parents are painters. I grew up surrounded by art and received a classical artistic education. After completing a graphic design school, I ventured into the world of video games, where I held various visual roles ranging from Concept artist to 2D Graphic designer, Animator, user acquisition/marketing Artist, and eventually I found myself unexpectedly in the role of Art director.
I love Art.
By the end of 2022 and into early 2023, our team took the decision to present a Demo of our game Little Legend at the Gamescom in Cologne. While under immense pressure, I took on the responsibility of painting and building the entire world of the Demo — a task that would typically be divided among five people:
- 2 environment artists to paint the sprites of the world and the background
- 1 character artist/animator to draw, rig and animate the characters and the monsters
- 2 level builders to place the 2D assets in Unity 3D, set up the lights and the VFX (fog, rays of lights…)
Here is a video showing the scale of the world we created in about 4 months:
My co-founder joked, saying:
"It would be funny if we used A.I. and finished the Demo thanks to it!"
…Little did he know…
As the Gamescom date approached, the booth was booked, train tickets were purchased, and with only 1 week left to finish everything, I still had 20% of the world to build!
Panic set in.
RED SQUARE represents 1 screen (1920x1080)
YELLOW CIRCLE represents what I had to create in 1 week
I knew that at the time A.I. could generate nice stuff, but it always looked kinda wrong to me — generated pictures felt soulless and I knew, most of generative A.I. models were training on artists work that they would find online...
This is a sad fact, sometimes a trained eye can even spot the exact references the A.I. used to generate it's result, no matter how hard it's creators try to hide it and social platforms lying about it.
Look at Pinterest, once a great place to find references, now a cesspool of cheaply generated trash…
These were my thoughts at the time…
and this is also when I decided to stop being negative!
Summoning my courage, I reorganized my workflow, changed my habits adapting my ambitions to reality. I started to feed the A.I. my own work to see how it would react, what did I have to lose?
Piece by piece, the A.I. was eating and regurgitating my stuff…
From there, it all clicked.
I started correcting the generated results.
I didn’t have to wait for the A.I. to give me perfect results, I just had to manipulate it enough to get small scale things perfect and then rework the little details.
After all, my art style is made out of little details!
Here is the result in-game:
Notice how my hand painted tree works well with the A.I. generated cliff edge?
And here is the biome created at 90% with “A.I. generated then overpainted sprites”:
The environment sprites are not as lush and colorful as in my other scenes, but they fulfill their purpose and I think it works well enough for a “dark cave”.
Eventually, I finished my work in time for the Gamescom and we received only positive feedback!
Weeks after the Gamescom, I did some new tests.
I decided to incorporate generative A.I. into my concepting process. Here again, I hit a “creative wall”, the results I got were good enough to pass as concept art, and sometimes as “finished assets”, but something was still missing.
Here is a very raw integration of my characters into A.I. generated scenes:
At first glance, it works well enough to pass as advanced concept art, but if you start looking closer, a lot of details stop making sense.
And unfortunately in a video game where the player explores and interacts with the World, we need more consistency.
Shapes need to have a purpose because the human brain can only be fooled for so long. Our eyes perceive that the A.I. doesn't comprehend the distinctions between interior and exterior, or the purpose of each individual element within the scene. When it encounters uncertainty, it tends to mix everything together, like a child smudging colors. Additionally, the A.I. struggles to follow instructions aimed at imbuing deeper meaning into designs or refining specific details as intended by the Art director.
At least for now…
Now, let's draw some conclusions.
A.I. is good at creating shapes but bad at understanding purpose. This is why I think a human artist expertise is needed.
Traditional artists and photographers panicked when Photoshop came out, some were saying that “photobashing was cheating” (a technique where artists merge and blend photographs or 3D assets together while painting and compositing them into one finished piece, speeding up their workflow to achieve a realistic style) and even today there are still Autodesk 3DsMax / Maya defenders saying “Blender is not up to industry standart” and yet, the software gains in popularity.
Paraphrasing Feng Zhu, founder of FZD School of Design and video game and cinema veteran designer:
Each time a new tool comes out, it disrupts the industry and it always goes through 3 stages:
1. That’s cheating
2. Everybody starts using it
3. If you are not using it, you are done.
I mean, looking at the past, he is not wrong!
So here is what I think:
A.I. is here to stay
A.I. is a tool on it’s way to find it’s niche
A.I. doesn’t know how to tell stories... and in today's world, stories, but more importantly, human presence, human touch, human emotions are what we humans need the most.
also…
I believe there are good uses of this tool that will become unavoidable and essential. A.I. is still in its infancy and, like a child; it needs guidance.
Now, what should I have done to make things the right way?
Not finishing my Demo and presenting an unfinished project at the Gamescom was out of question, I’m glad this “tool” helped me completing my work in time.
What do I think about artists being robbed or affected by A.I. ?
I guess if it happened to me I would have to adapt somehow like I always did.
What should I say to people who warn me that A.I. could steal my art for other users to use in their work?
I know that it won’t make me stop creating “traditional” art.
I am ready to hear opinions on the topic, but I fear they will come as polarizing.
Thank you
