william_kent

Well-known member
A thread for discussion and demonstration of algorithmic "art" - either procedurally generated or created by AI ( neural networks, machine learning, etc., )

Some examples of procedurally generated pictures:

Lingdong Huang is a bit of a genius - here's an example of "flowers in the style of a Chinese watercolour" - these can be generated on this webpage: nonflowers

1628865308258.png

Lingdong Huang's procedurally generated fish drawings: fishdraw

1628865496041.png

Lingdong Huang has also created another page which generates a "Procedurally-generated vector-format infinitely-scrolling Chinese landscape"

--------

The last couple of months has seen huge strides in the AI generated art world - this article, Alien Dreams, describes some of the latest developments and touches on some of the philosophical issues - can one take credit for content that has been generated by a computer, is the "artist" now an editor ( I refuse to use the "C" word here ), is it art if there is no human intent behind it?

I've been playing with this google codebook, VQGAN + CLIP ( Codebook Sampling ), where you can provide a text prompt, an image prompt, or both text & image. I've posted some of the results in the treeshagging thread, but to avoid cluttering that up, I'll post some here

Here's one inspired by a request from @WashYourHands - "All around the great oaks sang them to sleep and dreams" in the style of Max Ernst

1628866216877.png

All around the great oaks sang them to sleep and dreams

I've ended up generating quite a few based on WYH's choice of phrase + "in the style of Max Ernst" and other artists...

1628882736644.png

All around the great oaks sang them to sleep and dreams

1628866785896.png

All around the great oaks sang them to sleep and dreams

1628867500323.png

All around the great oaks sang them to sleep and dreams

1628867893769.png

All around the great oaks sang them to sleep and dreams

1628867941329.png

All around the great oaks sang them to sleep and dreams

1628868399184.png

All around the great oaks sang them to sleep and dreams

1628868547676.png

All around the great oaks sang them to sleep and dreams


-------

My personal rule is that there should be no "post-production" by humans

I suppose a real "digital artist" pipeline would be generation -> NFT, but why should anyone pay for AI art? People are charging for this stuff though...
 

version

Well-known member
That version of Napoleon Crossing the Alps I found a while back is still my favourite example of this sort of thing.

0-SSv-HWCj-AY-OALci2.jpg
 

WashYourHands

Cat Malogen
Just as the tincture kicks

I won’t cheapen the point with trippy, but that was an expeditionary quest, unexpected interior exterior interior worlds

These are insane, Ernst meets Nash

7A593ACB-F70D-46AF-BE92-57831B146A84.png

Leonora Carrington channeling

65A6FC1E-84F9-49C6-A7A7-D02CF50552AF.png

What a mission, thank you
 

william_kent

Well-known member
Just as the tincture kicks

I won’t cheapen the point with trippy, but that was an expeditionary quest, unexpected interior exterior interior worlds

These are insane, Ernst meets Nash

View attachment 7922

Leonora Carrington channeling

View attachment 7923

What a mission, thank you
well spotted on the Leonora Carrington!

bonus! here's another using the same prompt: "All around the great oaks sang them to sleep and dreams | in the style of Leonora Carrington"

1628895962689.png
 

Corpsey

bandz ahoy
It's presumably going to have a huge effect on art (if it isn't already) when machines can generate more original art than humans. The biggest effect since photography? (Except it will effect music and film and literature as well.)

If the world isn't on fire/underwater first.
 

william_kent

Well-known member
It's presumably going to have a huge effect on art (if it isn't already) when machines can generate more original art than humans. The biggest effect since photography? (Except it will effect music and film and literature as well.)

If the world isn't on fire/underwater first.

I've seen this "VQGAN + CLIP" AI generation referred to as the biggest development since Photoshop, but you've got a point that this may be the biggest effect since photography - very astute
 

william_kent

Well-known member
@linebaugh - spill the beans ( or "tea" - is that what the kids say nowadays? ) - how were your examples generated? was there any human intervention?

I'm curious...
 
Last edited:

Clinamenic

Binary & Tweed
Brilliant stuff, this is even better than the text-to-image stuff I mentioned a while back, more inspired, seems like it can hone itself more finely (and fantastically).

Are you doing this all manually, or is there some kind of application I can use online and/or download?
 

william_kent

Well-known member
Are you doing this all manually, or is there some kind of application I can use online and/or download?

I'm using this online codebook: VQGAN + CLIP ( Codebook Sampling ) - there are others but this is the one I've had the best results from

It is possible to run the code locally, but you'll need a good nVidia GPU, plenty of disk space, and many hours to download and install all the dependencies - it's just quicker and easier to use google collab and run the code on their infrastructure - you can get one to two hours of GPU use for free, but I found it was worth paying for the "pro" option to get access to the better GPUs and longer run times
 
Top