“magical wet-plate portrait photography, man crying, flowers, scratches to magical wet-plate portrait photography, woman crying, flowers, scratched” Stable Diffusion with Deforum Animation

After my initial foray into AI art and discovering Stable Diffusion I was totally hooked but also trying to find ways to use the technology to make something interesting to me.
It had become very clear it was easy to generate amazing looking images, particularly if you were prepared to put in some time learning the syntax and craft of prompt writing to sculpt the outputs into ever more beautiful/weird creations.

I needed a direction or at least some idea beyond just single beautiful images.
Then I discovered Deforum.

Put simply Deforum allows you to make animations with Stable Diffusion generated images. You can morph between prompts, add zooms or rotations and so much more.
It is, for someone who has spent most of their adult life playing with moving images in one form or another, insanely addictive and inspiring.

The first animation I created after installing and figuring out how to use it was the one at the head of this post.
As was becoming a daily occurrence whilst playing with this tech, I was blown away.
I just sat there gawping at the screen, once again struck by the insane possibilities of this cutting edge tech I could run at home on my slightly aging desktop PC – for FREE!

The next animation I made (below) I played with a little more, taking the output from Deforum into After Effects, tweaking and creating the reverse half of the loop.
Now we were getting somewhere! I could already think of a thousand use cases for this in my own work.
It also took me down a rabbit hole of ethical considerations……..

“0”: “an old man on a motorbike by mark rothko, Ewald Rübsamen”,
“20”: “a beautiful tree, trending on Artstation, by mark rothko, Ewald Rübsamen”,
“40”: “water by mark rothko, Ewald Rübsamen –neg photo, realistic”,
“60”: “the moon, trending on Artstation, by mark rothko, Ewald Rübsamen”
Stable Diffusion with Deforum Animation

You’ll notice in the prompts the names of two artists. Stable diffusion knows what those artists work looks like and so is able to generate images in their style.
I’m a huge fan of Mark Rothko, his room at the Tate Modern in London is one of my favorite places on earth. I chose to add his name as I wanted something with big blocks of colour, but not bright and garish.
Ewald Rübsamen, however, I had never heard of.
The “insert random artist” button in the Stable Diffusion web-ui added it for me.

Something felt, odd, wrong even about using a model trained on these artists work, without their permission, to do what I was doing.

But was it?

(Interestingly, I have discovered that the LAION model used to train Stable Diffusion contains at least 3 of my photographs (LINK) scraped from this blog all clearly marked as “©Uchujin-AdrianStorey” so it could be argued I have skin in this game)

The ethics of these models is a debate that is raging across the internet (Search it)
There are far more intelligent and erudite people than me discussing these issues and I would urge you to do some reading if this interests you at all.
I think it needs some serious thought by anyone looking to interact with these technologies.

My current feelings on it (which may change) are this:

– “Good artists borrow, great artists steal” a quote attributed to Picasso.
I know as a photographer and videographer that I have been heavily influenced by work I love, I have blatantly “stolen” ideas or techniques from others and integrated them into my own work.
Surely all artists do this? even if it’s subconsciously.
No one who is a painter/photographer/designer/musician has never seen/heard another painting/photograph/graphic design/piece of music!
And those works produced by others have influenced us all.
Isn’t this some form of the same?

-There seems to me to be a very real difference between playing with this technology to generate images/animations for social media (or this blog) with attribution if artists names were explicitly used and attempting to pass the work off as entirely our own, without reference to how it was made, and trying to monetise it.
One of those feels like artistic exploration the other feels blatantly wrong.

-Every time a new technology comes along it shakes up the status quo.
I’m old enough to remember when Digital SLR Cameras exploded into the world. Traditional film photographers were not happy.
The Canon 5D mk I was the single most important thing that made me start making films. Because suddenly I could. The technology to do so had been democratised in a way it had never been before.
Let’s not even get started on CDs, Photoshop, Smartphones or The Internet for fuck sake!
We will have to address how these models are trained, how attribution is given and have open conversations about what this all means, but the genie is well and truly out of the bottle and sat on your sofa with his/her feet up sipping a nice glass of red and has no intention of going back in.

It really does feel like we are witnessing something that is set to change the world in ways we cannot even begin to imagine yet, and much like “the internet” that will equally inspire, amaze and bring us together with one hand and disgust, oppress and divide us with the other.

We really should be used to it by now.

Part 3 of this mini blog series will deal with my narcissism :)