How to build an NFT series using AI (Part 3)
ICYMI: Chevy is feeding his original photos to an AI to create an NFT series (and documenting it all here). Follow on Twitter for first looks.
Ok, so I've had my suspicions…but this confirms it.
No matter how many of my own photos I feed Midjourney to use as references - none of the generated art is really mine.
Sure, according to Midjourney's terms and conditions, I legally own it - but to say I created it? That feels like a big ol’ stretch of the truth.
Here’s the thing though…
It ain’t Midjourney's either. Nor is it Larry Sultan’s or Petra Collins’.
It’s its own thing - its own 'entity' (?).
...which means it needs a name.
Preferably a dumb one - you know, to lighten the mood and help gloss over the fact that I'm using a civilization warping (potential world-ending) technology to make silly photographs in my spare time...
How’s ‘Billy Chadison’?
Dumb enough for you? Good. Now, go follow him on Instagram.
(Seriously, this is happening and you just need to embrace it).
Ok, but hold up. If this is now is “Billy’s” work:
Where do humans fit into this equation?
Fair question. If I had to define my role in the creative process, I'd give myself the title of ‘artistic director.’
I give Billy references, feed him notes, and request revisions until the work matches the theme I'm aiming for.
This is done by using a subset of reference images along with the same prompts with slight variations - which leads to each photo being different, while fitting an overarching style.
But just like yesterday - I'm out here telling you, when I should be showing you...
Here’s a visual rundown for you:
The prompt here was: ‘Nasa, mom, dad, kids, photo by Larry Sultan and Petra Collins.’
This result got me thinking:
"Huh, that’s cool.
Adding ‘Nasa’ to the text prompt put everyone in jumpsuits and added some space age looking fuel tank to the roof of the car...how do I get more of that??
...what if it was as simple as re-entering the same prompt + reference image? Surely not.
That'd be too easy...right?"
Apparently not.
Same prompt. Same reference photo. Different result (with a consistent style).
And the process of re-submitting the prompt and reference image to Midjourney took all of 2 seconds.
(There's literally a '🔄' button that you can click to resubmit your request).
To harp on my point from yesterday:
AI hyper contracts all of that time, work, learning, trial and error that exists between ‘idea’ and ‘final creation’.
I didn’t have to scout a new location, hire models, monitor the weather, set up any lighting, process any film, or change any lenses in order to experiment further here...
All it took was two clicks. (Spooky, no!?).
It feels like a creative cheat code...
So what’s the difference between building a photographic style manually, and doing it with AI?
(No surprises here) it’s different in almost every way. Every way, except for one.
I could be way off here, but this is how I'm starting to see it....
The process of creating your own art style is different in each situation, but the formula is the same. Here's what I mean...
The formula:
Personal taste + influences = new art style.
(The stronger your personal taste is, and the more influences you pull from, the more unique your style becomes).
This formula seems to stay the same, regardless of process.
Process vs. Process:
The manual process of running that formula (in the case of photography) might look like this:
Take in a range of influences, while testing different film stocks, digital sensors, formats, compositions, lenses, camera bodies, lighting effects, subject matter, set styling, chemical processing, digital post effects, etc.
Repeat that process over years of trial & error until you have yourself a new photographic style.
The AI process looks like this:
Supply the AI with your own original photos (informing your personal taste), describe the scene you want depicted and list your artistic influences.
Repeat that process with 5-10 mins of trial & error (tweaking prompts and reference images as you go) until you have yourself a new photographic style.
Ok…but where’s the line between influence and plagiarism?
There’s been push back from many artists, that have a very understandable (and justified) problem with AI models that allow individuals/companies mimic their artistic styles almost 1:1.
I'm not here to mimic, but to blend and create anew.
Which is a nice sentiment - but it doesn’t answer the question: where’s the line between influence and plagiarism!?
For me, I’m drawing the line using a simple question:
If [artist(s) you’re using as influences] were to put on an exhibition showing the AI based work you’ve created:
Would everyone be shocked by the new direction of their work and the apparent departure from any previous styles?
Or better yet: if the work didn’t have an artist's name credited next to it - would viewers be at a loss as to who was responsible for making it?
If yes, you’re good. It has become its own thing.
If no, it’s too close to that artist's style.
(It ain’t perfect - but it works).
Ok, that’s it for now - see you in Part 4.
Follow on Twitter and Instagram for first looks on each NFT release.
Want a chance to win a 1/1 NFT at release? Enter here.
P.S. Want to try making your own AI images?
Sign up to our newsletter and invite a friend using the referral code in the welcome email.
You’ll then be invited to our Discord, where you’ll get free access to Midjourney (the AI tool we’re using for this project).