AI generated art

[KGeo777]

That all being said, I love these images. Whether Fritz lang would have created anything similar is debateable (although I do think there is more than a hint of Metropolis here) but I would love to see a Dune movie with these kinds of set and costume designs - kind of Art-Deco Dune

A few of the images do make you go: wow--I can see this as a 1920s sci-fi movie. I especially like the image of the woman standing in the sand dune with the figure in a gold engraved mask.
But so many are retreads and it kind of gets sterile and icky.
It needs a guiding hand to make organic sense of the ideas.
It doesn't feel Fritzy Langy to me except maybe a couple of shots. It's too busy and repetitive.

 

[Daysman]

Still playing with Dalli2 by way of Bing - tried throwing descriptive parts of my 75 & 100 word tales at it. Results vary, but here's the Mage's daughter:

Note: the AI decided on a black and white shot.

 

[KGeo777]

I asked for Lovecraftian monsters and it knew what to do.

 

[KGeo777]

?

 

[KGeo777]

It is pretty amazing that you can type in "end of the world Jack Kirby style" and get something that really does get close to the style. The computer extrapolating from that quite accurately.

 

[KGeo777]

Midjourney - Vincent Price as the Grinch

 

[Harpo]

And now, Dall-E 3

 

[Rurherfurd Poma]

That kind of pictures are awkward, I don't know what's next the IA will do

 

[Daysman]

Discovered yesterday Bing's image generator is using Dali³ though I'm unsure if it's an official launch? At least one YouTuber claims it's not universal...

Anyhow, here's another starship, same prompt I've used before:

And here's a sff throne room with all kinds of stuff going on:

Oh, and the creature from one of my 75 word challenge entries...

 

[Rodders]

A friend of mine has been showing me AI generated art using Mid Journey, which has been hugely impressive.

I am currently in the camp that doesn't accept AI generated art as genuine art and I wonder whether this is a new form of prejudice

 

[Robert Zwilling]

Not liking something might or might not be be prejudice. It all depends on why it isn't liked.

Modern art, especially conceptual art, is disliked by a number of people and not considered to be art.

Is this art?

AI art will probably become just another facet of conceptual art. Eventually 3D sculptures from 3D printers designed by AI programs will also become common place.

AI has already steam rolled over any objections that could stop it from becoming a physical reality. The public has accepted it as art, certainly as something to appreciate and also buy. The only question is whether it is art or not, like many other forms of art, that question is answered by explanations that only leave more questions than answers.

One real question is will it matter how the original programs were developed using other peoples work without their permission. That's a legitimate reason not to support something. Even if the courts say that the original work of others can't be used, the programs are already made, they can now work off of anything, so the programs aren't going to disappear. Just another example of computer assisted living.

 

[M. Robert Gibson]

I've been reading into Stable Diffusion, and one section about tokens caught my eye

How does Stable Diffusion work?

Stable Diffusion is a latent diffusion model that generates AI images from text. Instead of operating in the high-dimensional image space, it first compresses the image into the latent space.

stable-diffusion-art.com

"Stable Diffusion model is limited to using 75 tokens in a prompt. (Now you know it is not the same as 75 words!)"

75? That number seems familiar. Aha the 75 word challenge! So I thought I'd use my October's effort as the prompt

Here is the result without the title

And here's the result with the title, using the same seed ( I simply used the id of the challenge thread: 585731)

 

[Stephen Palmer]

Made with Midjourney.

 

[Mgellis]

I don't think purely AI-generated art can be recognized as "art," in the way that the Mona Lisa is art. It is not something made by a person. Too much human creativity has been taken out of the process. Just because I typed in a prompt does not mean I can take credit for the image generated by the AI.

AI-assisted art? Where the artist does most of the work and the AI does things like smooth out rough edges, etc. is a big gray area. Yes, it means you need less technical expertise to create fantastic images, but the image is still primarily the result of human creativity. It's like using an image editor to make an artistic photograph better-looking. But how much is too much?

Related issue...AI art for commercial purposes. Yes, it means a person without artistic ability can create lots of pictures for book covers, etc., but it makes it harder for artists to sell their own original work. Of course, whether we like it or not, that ship may have already sailed on the winds of economy.

The real value of AI art, I think, is as a brainstorming tool. It's amazing, for example, what you can learn about a character by having an AI create different versions of that person, which it can do very quickly, and all you have to do is type in prompts like "Short and broad, with tan skin and delicate features; this person has shoulder-length curly dark brown hair with thick body hair, deep-set gray eyes with thin eyebrows, a wide, smiling mouth, a bulbous nose, and pointed ears."

I hope this helps.

 

[HareBrain]

Made with Midjourney.

View attachment 112809

This is one of the first pieces of AI art that I could have believed was made by a human (until you look at the reflections, which are wrong), because I could see why a human artist might put the effort into making it. Pretty much everything else looks impressive at first glance, but the amount of effort it would take a human artist to make is at odds with the fact that they are so boring. They say nothing. The faces depicted are almost all completely lifeless.

This might change, but as the amount of art used to train AI increasingly includes earlier AI-generated images, it could well regress.

 

[Stephen Palmer]

This might change, but as the amount of art used to train AI increasingly includes earlier AI-generated images, it could well regress.

There's an interesting hypothesis going around that says this is happening already.