|
Post by BlueAndGold on Jun 15, 2023 12:18:33 GMT
I'm curious if anyone here has experimented with AI Art generators to help illustrate their books or create covers, and what their experiences have been.
|
|
|
Post by Mag2024 on Jun 15, 2023 12:44:31 GMT
I'm curious if anyone here has experimented with AI Art generators to help illustrate their books or create covers, and what their experiences have been.
I've played with it for story. It was predictable and lame. No imagination. I've heard they are great for art. A million things you can do. You do not own copyright, however.
|
|
|
Post by Retread-Retired-Cameron on Jun 16, 2023 14:27:41 GMT
I tried it a few times and the results weren't great, unless you're illustrating a horror story. The extra fingers and at time appendages weren't exactly inspiring. Sometimes the faces came out well, but most of the time they didn't.
|
|
|
Post by BlueAndGold on Jun 16, 2023 16:45:56 GMT
I've been playing a bit with two online AI generators thinking I might use them to illustrate a book, but my experience is the same as yours, Cameron. The AI can create some supremely beautiful things - if you leave it alone and just be happy with whatever it makes. But if you want to direct it with text to create your vision, it is TERRIBLE at following directions! Landscapes and buildings aren't bad after numerous attempts, but human figures are almost invariably horribly twisted and deformed: People sharing limbs, multiple mouths on faces, heads on backwards, too many limbs almost every time, too many people (ask for two and your get three or four, usually sharing body parts), limbs morphed into background or foreground objects, horses with the heads of different animals. I'm sure there are operating principles I'm not aware of, but the sponsors of the two generators I have been playing with don't make them obvious. Very disappointing! As Maggie pointed out too, it has still not been decided in the courts just who (or what) the copyright of any particular produced image belongs too; there are still active cases in the courts. Troublesome! I guess I need to learn to draw again...
|
|
|
Post by Retread-Retired-Cameron on Jun 16, 2023 17:51:44 GMT
Images you make with programs like MakeHuman and Blender can be Copyrighted as IP, same as images you make with other software where you are the artists. The results depend on your skill level and talent as an artist using a graphic tool.
The problem with AI art generators is figuring out how to format the text to get the program to 'see' what you want it to output.
|
|
|
Post by And still Kevin 2024 on Jun 16, 2023 23:18:50 GMT
www.makeuseof.com/copyright-rules-ai-art/The point arguing against copyright in that article that nothing is original, is only partly true. It's like saying books cannot be copyrighted because they use the Alphabet. I read a lot of fiction and often think, hrrm this plot sounds so familiar. It's hard nowadays to be 100% original, but that does not imply plagiarism. But no I have never used AI to create my covers or specific art, I doubt it could. But why would I want it to? Blender and software like it. Yes. As long as what you use them to create is original, enough, they can be copyrighted. But they are not AI. interestingengineering.com/culture/what-is-ai-generated-artAll my art nowadays is created freehand on my PC. It only exists as jpgs. Around five years ago I contacted a gallery in London, who I know sell paintings, and prints of them, asking if they were interested in selling my art. When I told them how it is created, they said, but it's not art! I wonder what their opinion is of David Hockney? Granted he started off with actual paints, but now uses Photoshop and an Ipad to create art. One of his paintings sold a few years ago for $90.000.000. He's not as yet moved on to AI to create art for him. Also not AI art, but art by a monkey. 55 paintings put on display. Sold for an average $10,000 each. All understandably totally original, but give AI a few years and it will churn out original art if instructed to, and I have no doubt people will buy it. Just like people now pay millions for art done using stencils by a 'well-known' graffiti artist, who apparently cannot copyright it, otherwise he/she would have to sign his/her name on it. Although there's a lot of classic art that predates copyrights!
|
|
|
Post by Mag2024 on Jun 17, 2023 13:37:21 GMT
I've been playing a bit with two online AI generators thinking I might use them to illustrate a book, but my experience is the same as yours, Cameron. The AI can create some supremely beautiful things - if you leave it alone and just be happy with whatever it makes. But if you want to direct it with text to create your vision, it is TERRIBLE at following directions! Landscapes and buildings aren't bad after numerous attempts, but human figures are almost invariably horribly twisted and deformed: People sharing limbs, multiple mouths on faces, heads on backwards, too many limbs almost every time, too many people (ask for two and your get three or four, usually sharing body parts), limbs morphed into background or foreground objects, horses with the heads of different animals. I'm sure there are operating principles I'm not aware of, but the sponsors of the two generators I have been playing with don't make them obvious. Very disappointing! As Maggie pointed out too, it has still not been decided in the courts just who (or what) the copyright of any particular produced image belongs too; there are still active cases in the courts. Troublesome! I guess I need to learn to draw again... Find an image you like and use it as a template. Work over it. Good luck.
|
|
|
Post by And still Kevin 2024 on Jun 17, 2023 23:29:31 GMT
Current AI has no intuition. To get it to create something specific you have to give it precise instructions. Such as, analyse all of Rembrandt's art and then emulate it with an original image in his style. Although you would then have to build a robot that could paint it, not print it, paint it. www.bbc.com/future/article/20221123-the-weird-and-wonderful-art-created-when-ai-and-humans-uniteSome of that is really an example of not fully knowing what AI is. The part about the fish is simply a type of logic algorithm from decades ago, It was used in text only adventure games on computers. It's a complex form of search engine with a large database (not WWW to search in those days!) It was able, usually, to read up to four words in a command. Such as "break door down." But it was complex enough to know there were many options, but only one way to do it. "Break door down with what"? "break door down with brick". It would search if you had a brick, but even if you had, it would often say "you cannot do that" because the answer would be programmed in, and that's not it. "Break door down with axe". It would look if you had found one, look if that was the solution, and tell you the door had been broken down. If you did not give a precise solution it would refuse to do it. "Smash door down with axe" would not work because it wanted the word, break. Bolien logic I think it was called. Later used by Ask Jeeves, and then much later by some search engines, but many still seem only able to 'read' and 'understand' the first two words. It's not AI. True AI would be able to intuitively work out what you mean, even from an entire sentence, and come up with the correct answer, like a human. Like some anyway! I don't know if you have seen Blade Runner, but the main character in that hunts down artificial humans and terminates them. But they are so perfectly created that the only way to tell if they are not human is for an expert to do a very complex Turing test interview. I would have thought if they are so hard to tell apart from humans, outside and in and intellectually, true AI, why did they not just leave them alone?
|
|