How I used AI to generate an NFL Helmet Design
I’ve been experimenting with AI art recently, a subject I’ve been fascinated by for the past 2 or 3 years, mostly as an observer. Artificial Intelligence is, I believe, the technology of my lifetime and an oft-fantasized subject in the Sci-Fi and Fantasy stories I’ve loved since childhood, so it’s incredibly exciting to see this come to pass in my lifetime.
One of these experiments was to try to recreate 1960’s Hannah-Barbara style animation logos for various NFL teams — inspired by the infamous “Brownie” logo of the Cleveland Browns.
The first team I tried was of course my hometown Seattle Seahawks. [Note — all generations are via Midjourney]
Because AI is referential rather than truly imaginative, I asked it to come up with a logo for the fictional “Thunderbirds” football team so that it wouldn’t simply try to recreate the real Seahawks look. Intrigued by the 3rd option (the aqua blue helmet with the feather-mask look) I asked the AI (Midjourney) to come up with some variations on that idea, but as a helmet design rather than a logo:
Whoa. Those look really good! Sure — the helmets are a little bit off and the logos aren’t quite right but — just really liking the concepts and gestures.
I’m going nuts at this point — with a little cleaning up and editing these could be some really cool looking helmets; I could slap the design on one of the templates that floats around the Chris Creamer Message Boards (the volume of which has clearly influenced the AI results). However, due to my excitement, I started making conceptual designs for all the teams in the NFL
As you can see, there are just a ton of great concepts here. But the AI doesn’t always come up with ideas that I find compelling — I had some trouble with coming up with a design for the Jacksonville Jaguars that didn’t look too much like the existing look —
None of these felt really compelling — perhaps because ‘Stalkers’ was a poor substitute for jaguars but unfortunately, Lions, Bengals, and Panthers are already NFL teams, so any of those substitutes might hew too closely to existing designs. But I hit the “reroll” button a few times until I came up with these:
That one on the bottom right intrigued me — I liked that it looked like the leaves in a jungle (from behind which a jaguar might be keenly observing you). I asked for some variants:
We’re getting somewhere! I’m into this blue and gold jungle kick
Even despite the Carolina Panther peaking out from the version on the top right, I really liked the way the leaves overlapped on that design. It intimated a lush jungle. It didn’t look like any helmet I’d ever seen either, so I decided to go with it. I like that it’s non-traditional, even by the standards of the Jaguars who in the past have graced us with the following disaster:
Anyhow, in order to turn this into a real design, I’d need to come up with a flat version of those overlapping leaves and fronds and such, and then use a Photoshop template to render it onto an accurate looking template. We’re going right back to midjourney for that.
These look great. I picked one and upscaled it, then hunted for a suitable .PSD template. I wanted to get one that I saw frequently on Chris Creamer / Uniwatch.com redesign contests and I found this-
This isn’t an AI design, someone at some point designed this to make helmet mock-ups and it’s been passed around the internet for ages. All I’d have to do is adjust the colors of various bits, and the template would take care of making it look like a proper helmet. I pasted in my design — easy because it’s just a simple square.
And there you have it! While the design may not be your cup of tea, I dig it. I can only imagine what a matching jersey set might look like (ed note: Should I ask midjourney to imagine it for me?). There are other Photoshop templates that show the helmets from multiple angles, and I’ll eventually find one of those to give this a try on, but this is good enough for now — I could spend forever tweaking the design and this particular template makes it difficult to get a nice looking helmet stripe. We’ll call it a day for now.
Further Thoughts:
All the above is basically something I do for fun, and doesn’t have much meaning in the real world, but nonetheless I think about the ethics and etiquette of using AI tools in this way. While I view much of the negative opinions of AI-art tools to be FUD/Ludditism, ultimately it’s public acceptance of the validity of the idea that’s more important than the specific ethics, which I liken to a camera (“all you have to do is twiddle with some settings and click a button — anyone can do that” — try telling that to Ansel Adams).
It probably took me 15 minutes of fiddling with the prompts to get to the design concept I ultimately selected; it took another 30 minutes to find and download the Photoshop template and perhaps another 15–30 minutes to work with the template and set the colors how I wanted them. The most time-intensive tasks were writing this essay, followed by the time I spent hunting for the photoshop template.
To me, the AI assisted me in two clear ways: While “the jungle” and the specific colors were my input and my, uh, inspiration, the ability to visualize that inspiration was critical. It presented me with 16–20 options in varying styles (of which my guidance — a specific decade and some colors — was extremely limited) when I may have only seriously pondered 2 or 3 in my mind; and I’d have struggled to find the words to coalesce the thoughts into mental images that I then could execute on-screen; instead I saw 20 ideas, all instantly rendered in something vaguely like it’s final form. Thought prototyping.
Secondly, it made decisions for me that were 1) trivial but 2) time consuming. Could I have found or designed a collection of jungle leaves myself? Yes absolutely. Could I have randomly tossed them down in some pattern on the helmet? Again, yes. However — and I’m speaking personally here — designing randomness is something that terrifies and paralyzes me. There are maybe 20 individual leaves on the final design, each at some point and rotated in some direction, in some stacking order. These used to be specific choices that I had to make and I’ve been stuck agonizing over — how dense should the leaves be? How random should the layout appear? Does it look random enough, or does it look “too random”? Is it uneven and clumpy?
These are questions I didn’t ask of the AI — I took what it gave me and moved forward. There’s perhaps a lesson in that, but it was convenient to outsource those decision.
Ultimately, the entire project took me under an hour when it would have taken me maybe 4–8 hours previously; if they had a classical logo I might have spent another hour or three refining what the AI gave me (which is something I’ll do my next project) before slapping it on a helmet design. And of course, it’s speculative that I would have had the inspiration at all.
So, what do you think? Did I cheat or steal? Is this an appropriate way for artists (I’m not self-nominating here) to create graphic design?
Lastly, I finally did generate that cartoon Seahawks mascot/logo:
