What Stable Diffusion thinks of the text input “Elon musk on the mars dollar bill, photorealistic, cinematic lighting, hd, high details”.
AI meets art and the disruption begins
“And now for something completely different” Fintech-inspired AI art brought to you by DiffusionBee and the Stable Diffusion. Stable Diffusion is a machine learning, text-to-image model developed by StabilityAI, in collaboration with EleutherAI and LAION, that generates digital images from natural language descriptions. A true text to image machine learning AI.
I’m known for keeping “on topic” and confining myself to fintech, China and CBDC but for this article I thought I’d try something a bit different and venture into the world of AI-generated art.
I’m doing this in part because I have a never-ending need for graphics to feed my newsletters and it’s not always easy to find images.
As many of you know, I’ve gone with classic paintings and images from famous artists for most of my recent newsletters because for captivating images you can’t do much better. But let’s face it, it’s hard to find a Visa card advertisement with flames in the background on a “Google Culture” or image search. If you do find one it’s likely you can’t use it due to copyright. Now I can make my own.
Input text: Visa credit card advertisement with flames in the background, hyper realistic, 8 k, denoised, cinematic lighting
Set-up for Mac only
To download the software head over to diffusionbee.com. For now the software is only compatible with Macs running the newer M1 or M2 chips and you must have the latest Mac OS. The software is resource intensive and on the latest MB pro with 16 gigs of memory images may take a few minutes. M1 Macbook Airs should run the software though images may take a while to render.
Stable Diffusion AI allows for text to image conversion which is a lot more complicated than it sounds as you have to be very specific in telling the AI what you want. Starting with simple text alone may not result in a striking image. To help visit Lexica.art for a catalog of Stable Diffusion images made on their cloud. By following a few of the previously used text strings you can get some astounding images.
Input text: Visa credit card advertisement with green alien in background, hyper realistic, 8 k, denoised, cinematic lighting
What is clear is that this is a tool, that has to be learned. The image you get is highly dependent on what you tell the AI and frankly the long text strings I’ve found on Lexica are a huge help.
For those who can’t run this on your computer fear not! There is a web-browser based version, go to:
Sign-up and you’ll be able to generate quite a few images before being asked to pay.
What about the artists?
The problem however is that some of these text strings reference living artists. When an AI copies a living artist in my humble opinion the artist has a right to be upset as the AI is stealing their intellectual product. This is just the beginning. I can see the law suits coming.
Note the artwork below copies the art of deceased Gustave Dore and Monet. Neither will complain. Living artists or the estates of more recently deceased artists may have other ideas. Take for example this image in the style of Roy Lichtenstein who turned 1950’s comic strip depictions into fine art. This image below is truly remarkable!
Input text: “A portrait of a woman holding a single golden bitcoin coin in her hand, 1950's, punk by roy lichtenstein”
How smart is it?
Now to fully grasp the brilliance of this AI it is important to understand that once you download the 4gb of libraries to your computer you have full data privacy - nothing is sent to the cloud! All of the thinking to generate an image is done on your computer. The amazing thing is that the library isn’t pictures per se, but the knowledge of the pictures. So when I ask for an image that copies Monet, the AI knows what Monet paintings look like. The depth of the AI’s intelligence is nothing short of a miracle and its easy to see why commercial artists loathe this technology. So when an AI generated image won first prize in a Colorado art fair this past week actual human artists were livid. I feel for them!
Input text: shanghai skyline by claude monet
Note that the buildings are fairly accurate.
Evolution of an image:
Take this image with a Chinese dragon sitting on dollar bills, an image that I’m sure you can all imagine me using someday!
Observe how the image changes as the description becomes detailed.
Input text: Giant Chinese dragon sitting on a large stack of US dollar bills.
The image works and for a first pass isn’t bad.
Now lets try again with a better description. I generated about 10 images before this one emerged, each image was very different. This one is more detailed and almost 3D in its rendition. I confess I like the first version better for newsletters but can appreciate the refinement in the second.
Input text: “Giant Chinese dragon sitting on a large stack of US dollar bills with a money band, beautifully, fantasy, dramatic, intricate, elegant, highly detailed, digital painting, artstation, concept art, smooth, sharp focus, illustration, art by Gustave Dore”
Note that this dragon is modeled after Gustave Dore etchings which is why it is rendered in black and white.
Final thoughts
As with all technology this is already being used for nefarious purposes. While the software will block nude images and vulgarity it is clear that this technology can still be used to generate violent or sexual images.
The number of “Elon Musk” dollar bills that I found online shows that it is already in use with celebrities. I had to make one of my own! What happens when these images are used to bully or depict fake news? We don’t know yet, but you can bet that these images will end up in the courts sometime soon.
This is fabulous technology and I hope you are able to run the software or enjoy the image library on lexica.art. I’m sure I’ll put this software to work in this weekend’s newsletter. I don’t see these images replacing the classic art I currently use but who knows when a flaming Visa card will come in handy?
For now, let’s take a moment to marvel at how truly amazing this AI is before we watch the court battles unfold.
This AI is just a small window onto an AI future that is coming for us all!
Input text: Dragon and the US dollar
Input text: Tech and banks
Input text: Red color flying dragon. Now the problem with this image is that it was based on a copyrighted image that I was able to upload and use as a starting point for this one. The problem is evident! And it didn’t seem to understand “red” dragon!
Thank you for reading!
You can’t possibly say no to Spock!
Hey did I mention that subscribing is a great way to say thanks? Every new subscriber helps me get my message out to more people!
More of my writing, podcasts, and media appearances here on RichTurrin.com
Rich Turrin is the international best-selling author of "Cashless - China's Digital Currency Revolution" and "Innovation Lab Excellence." He is an Onalytica Top 100 Fintech Influencer and an award-winning executive previously heading fintech teams at IBM following a twenty-year career in investment banking. Living in Shanghai for the last decade, Rich experienced China going cashless first-hand. Rich is an independent consultant whose views on China's astounding fintech developments are widely sought by international media and private clients.
