FA+
Shop
Is this the end of Mister Foxy?
a year ago
Hello to all my friends, fans, and all the wonderful, highly creative people here on FA.
I only see this new fad of AI art as a tool, or an aid to help with the creative process, and not an "End All" for creating artwork in its entirety. Yesterday though, I found out that Deviant Art is offering a new folder-page for AI art created solely through the new medium, and my curiosity is piqued about it. Of course, I plan to research the new DA page a bit further, once I get the opportunity to do so.
What I understand of it is that one can type in what they'd like to see into the program, and the AI searches the millions upon millions of examples of artwork, photograph, drawings, and other visual images for what it feels is close to the typed in request. It then "Doctors" up the images to closer fit the typed in request and presents it to the artist/user. I feel it's up to the user to use Photoshop, or some other image manipulation program to tighten up the image, especially in fixing odd perspectives and of course, those "Flubbed Up" fingers and hands.
I have yet to use AI to help my artistic process, and I'm really not interested in doing so. Still, I wonder if my current state of artistic depression might be brought on by the fact that my art will never be close to the "Perfection" I have seen other artists achieve, which effectively has created a new fan viewership all its own.
I can't help but be reminded of the model building revolution of the late 1980s where machines were created to build and paint diecast models by the hundreds, and some had fidelity and detail almost rivaling the winners seen at IPMS shows of those days. One didn't have to go through the trouble of building and painting a model kit, and some could even buy "Ready for the shelf" dioramas in a variety of subjects and scales, providing they had the cash to buy these high priced, so called, "Collectables."
Then, not too long ago, there was concerns that 3-D printing could take the charm and pleasures out of the model building hobby. until hobbyists accepted it as another tool to help with the model building and creative process. Personally, I LIKE the 3-D printing and rendering programs, especially in creating fantasy figures and machine components unavailable in the popular markets. Several of the friends I follow here on FA use Hero Forge effectively and believe me. If there was a 3-D printing firm that offered to create miniatures of my characters in 1:35th through to 1:12th scale at a reasonable price, by golly, I'd use their services!
When I first came into the fandom, Pagers were a big thing. They were quickly eclipsed by the flip phone, and in less than 30 years the Smartphone has become a "Standard" in the high-tech world. I have created art on a Tablet, and I found it good, but it has its limitations. I still can't get over the practice of using pencils, paper, pens and other "Traditional" means to create my images. But after all, I'm not in this for the money. My art is, and shall always remain my "Hobby," which I am happy to share with all of you following me here on FA.
"Peace."
I only see this new fad of AI art as a tool, or an aid to help with the creative process, and not an "End All" for creating artwork in its entirety. Yesterday though, I found out that Deviant Art is offering a new folder-page for AI art created solely through the new medium, and my curiosity is piqued about it. Of course, I plan to research the new DA page a bit further, once I get the opportunity to do so.
What I understand of it is that one can type in what they'd like to see into the program, and the AI searches the millions upon millions of examples of artwork, photograph, drawings, and other visual images for what it feels is close to the typed in request. It then "Doctors" up the images to closer fit the typed in request and presents it to the artist/user. I feel it's up to the user to use Photoshop, or some other image manipulation program to tighten up the image, especially in fixing odd perspectives and of course, those "Flubbed Up" fingers and hands.
I have yet to use AI to help my artistic process, and I'm really not interested in doing so. Still, I wonder if my current state of artistic depression might be brought on by the fact that my art will never be close to the "Perfection" I have seen other artists achieve, which effectively has created a new fan viewership all its own.
I can't help but be reminded of the model building revolution of the late 1980s where machines were created to build and paint diecast models by the hundreds, and some had fidelity and detail almost rivaling the winners seen at IPMS shows of those days. One didn't have to go through the trouble of building and painting a model kit, and some could even buy "Ready for the shelf" dioramas in a variety of subjects and scales, providing they had the cash to buy these high priced, so called, "Collectables."
Then, not too long ago, there was concerns that 3-D printing could take the charm and pleasures out of the model building hobby. until hobbyists accepted it as another tool to help with the model building and creative process. Personally, I LIKE the 3-D printing and rendering programs, especially in creating fantasy figures and machine components unavailable in the popular markets. Several of the friends I follow here on FA use Hero Forge effectively and believe me. If there was a 3-D printing firm that offered to create miniatures of my characters in 1:35th through to 1:12th scale at a reasonable price, by golly, I'd use their services!
When I first came into the fandom, Pagers were a big thing. They were quickly eclipsed by the flip phone, and in less than 30 years the Smartphone has become a "Standard" in the high-tech world. I have created art on a Tablet, and I found it good, but it has its limitations. I still can't get over the practice of using pencils, paper, pens and other "Traditional" means to create my images. But after all, I'm not in this for the money. My art is, and shall always remain my "Hobby," which I am happy to share with all of you following me here on FA.
"Peace."
But it has limits. A lot of them if you look into it. Mostly, it still can't think for you or come up with ideas or designs. While it can copy a known style, it can't come up with a style of it's own.
If you can put it into words and the AI has enough references it's likely it can turn it into a picture.
But then again even if it has references, they might not be what you want. Every time I wanted the picture of a unicorn, I got it in 'My little Pony' style, because that's the most references the AI has.
It gets even further. What I have seen of current Science Fiction I'd say that these 'professionals' have forgotten how to show us spaceships in cool shots. Short Fan movies still get it though. But it humans don't get it any more, how will an AI? And mind we are talking about humans that are supposed to create something people would want to buy merchandise for.
What artist might have to do is keep up with the limits of AI and learn to do art beyond simple Pin-Ups.
So, no AI will ever match your style, ideas or jokes and there will always be people that appreciate both.
https://www.deviantart.com/cheetaur
Incidentally, I find the state of AI art to be lacking in humanity, and to be somewhat creepy. But having a machine that "draws" is every exectutroid's wet dream. No more pesky artists to deal with. Some of the younger people I know have tried to incorporate AI into their own art, and it always sticks out like the proverbial sore thumb. It's at once astonishing, and disturbing. I've no interest in using it myself.
Sorry for the tirade, but it's past my bedtime.
But hey, we'll see.
Now, the future is always changing. And quite rapidly. For example, this Udio AI that does the audio stuff has been quite the opposite. A couple style quotes and a couple sentences... I'd swear the darn thing is reading my mind. Like to the point that for my projects, I may be willing to pay for it. The art scene, though... that's a waste of money... so far. Now will I feel that way come next week... or three? Who knows. But, right now, I'm still in the opinion that a Human artist understands what a computer attempts to comprehend.
Even DA, who claim to get their source material “ethically,” initially buried the opt-out button so that you’d have to individually edit every post you made and select that button if you didn’t want it to be used by their LLM. It was only after the CEO got booed on stage that they made opt-out the default.
Besides, our own little defects is also what makes the art we produce, for hobby or as a job, unique and beautiful in its own way.
That said, while there is always room for improvement, I personally wouldn't change a thing about your art.
Another personal note is that for how cool digital can be I always had a soft spot for traditional works, especially ink ones, and I thank you for sharing it with us here on FA or a website of your own (That's where I first found your art back in 2006).
Think about the way autocorrect annoys us. This is the same thing, writ large. Autocorrect annoys us because it lacks context and discernment and makes us write what we do not mean.
In the end the people who create the "AI" algorithms that do not have actual intelligence, they loathe the people who create...especially artists. They hate the demands and desires for creative people to receive compensation for the work. The idea behind this is to get the creation for free and show these artists and writers that they are useless.
Now many of these algorithms are "free," because they grab the effort of others to create the original seeds. But know that it's very computationally expensive to operate these algorithms, and store the data. Eventually, once it has enough seed data, it will have enough vendor lock-in to be closed off and monetized. Like so many of these fresh, shiny Internet technologies *Crypto, NFTs), it's designed as a free money printer for those who get in early.
The long and short of it is, keep drawing...what you do is special and you can't get your art out of a computer.
There are some things we humans have that is natural. We have an intelligence that can delve much more deeply into our creativity than AI ever can or will.
As long as we can engage our own imaginations in everything we create, AI will never touch us.
Ironically, the latter is what real people are doing with "inpainting", where you take an existing image, cut out a portion of it, and ask an art algorithm to fill in the gaps with what pixels it determines ought to fit.
Many things that may seem common sense to a traditional artist will be novel concepts to those who never took that path.
There is no replacement for this.
"What I understand of it is that one can type in what they'd like to see into the program, and the AI searches the millions upon millions of examples of artwork, photograph, drawings, and other visual images for what it feels is close to the typed in request. It then "Doctors" up the images to closer fit the typed in request and presents it to the artist/user."
I want to address this, because I think a lot of people think this is how AI image gen works, and it isn't, and there's a lot of mis-directed anger as a result.
I think the confusion comes when people hear about the training process, where an AI model is trained on vast data sets of images. They connect the dots in their head and go "oh, it's creating a library, and then combining from that library at the point of creation, like a collage but infinitely more complex because computer.
No.
This is explicitly NOT how neural net models work, and it's the thing that makes them impressive. The datasets these models are trained on are IMMENSE. If the model contained the training data, it would be enormous, and useless, because models have to be loaded into memory. The scale would be completely impractical. Actual AI models are a tenth, a hundredth the size of their training datasets, and CONTAIN NONE OF THE ORIGINAL DATA.
"Training" a model doesn't mean taking the training dataset and indexing it, or adding some meta-data or whatever. It means TRAINING. The model discovers, via painful trial and error, what a cat is, and how it's different from a dog. What "standing" is. What makes one artist's style different from another. It internalizes all these things, things that can't be explicitly and clearly described with human language. And then, when you ask it to draw a picture of a standing cat in this or that style, it draws on what it knows to make something NEW, from SCRATCH.
Human artists don't go around with massive catalogs of every piece of art they've ever seen in their heads. They study, learn bits and pieces here and there, and then use that accumulated, subtle understanding to make new stuff. Neural Net models work exactly the same way, just far less efficiently (so far). THAT'S what makes them so impressive, and so scary.
This is an important distinction, because it's at the crux of what makes this new AI stuff so powerful. They're not just indexing and regurgitating information, they're actually UNDERSTANDING it at some level, which allows them to be much, much more efficient in terms of size and processing. In the early days of self-driving it was a hopeless problem, because everyone was trying to explicitly program ever situation, every "rule" of driving, which is near impossible and massively expensive because of the compute and hardware required. Then (several years ago now) Nvidia had some people drive around capturing hours and hours of driving footage from two mediocre cameras, one front-facing, one back, and fed it into a neural net. Once the model was trained, it was small enough to run on a half-rate laptop, and using nothing but those two mediocre cameras, the model was able to drive the car reasonably well. Still far from perfect of course, but leaps and bounds beyond what the manually-programmed stuff had done, and for a fraction of the cost.
All of this to say, in short: AI models don't create by crude recombination. They create by using previously learned knowledge to synthesize something new. Of course, at a certain point you're just playing with words: even humans, in some sense, are just "recombining", taking all the half-remembered things they've seen and using bits from them to make something new. But at a certain point you cross over from collage or tracing to actual new work, and AI has certain done so.
Nothing can ever replace your art. You have a style and charm all your own, and no other artist can replace that even if they better in some way. Certainly AI can't. I'm honestly of the opinion that AI will never really be able to get certain thigns right because it's just a machine, and I can't imagine an artist enjoying "touching up" an image that was mahcine generated, any more than i would enjoy editing a story spat out by a machine.
I'm with you when it comes to working with traditional drafting and drawing materials... It's hard to beat a pencil and paper. When I was learing CAD I found it so damn frustrating because nothing was intuitive.
You do you SteamFox.
Some were a flash in the pan. Some became pervasive tools. None of them killed anything people thought they would.
I think that AI will lower the bar for creating mediocre content. Kind of like 'Bucket Fill' lowered the bar for creating mediocre art as compared to learning proper use of physical media. Just like Photoshop - and Krita and other digital art tools - can be used poorly or used well, I suspect that AI will stay in a state that can be used poorly or used well.
Examples:
Cooking used to be a life skill, now it's either a hobby or a profession and not covered in a general education. (Then again, civics seems to be gone from gen-ed, too.) There's still makers of buggy whips, but they're not common or industrial and very little effort is spent trying to improve their designs or materials.
(Not to mention AI almost always botches the hands, feet, eyes, ect... 😝)
As for 3D Printing, someone still has to take time to model the files - it honestly can take as long - if not longer in some cases - to render a 3D Model (unless it's 3D Scanned, though that's more or less like using an old Pantagraph Machine - just a bit smaller, and more versatile today), and get it ready to print... Plus you still have to paint & detail the finished project! 🖌😉
I find it's a bit like throwing darts, you kinda get what you want. The AI starts with random static and attempts to turn it into things it recognizes from the keywords. Negative keywords sometimes have unintended effects, like 'no blue cows' means 'oh, you WANT COWS, just not blue ones.' Likewise, Thunder-Thighs got me a lot of storm backgrounds, and Mountain Lions always came with mountains in the background.
There is a way, if you set Stable Diffusion or perhaps other engines, up on your machine to get it to load your art and 'finish it' in a style. I've seen several people do rough sketches and have a Marvel comic books style pass that produces something reasonable. Youtuber Shadaversity, and FA Artist Icon:furronika have done things like that.
I have a DA account, and have been spamming the void with a bunch of AI things there. If you look, sasta-koogrr isn't too hard to find. I've done a lot of experiments, with a friend, of running the same prompt on different engines to see what it does. Some are better than others, some CLEARLY looked at a lot of furry art.
If you want, note me and I'll give you a Discord or Telegram link and I can share my experiences and some links.
Overall I agree, AI is good for prototyping, but sometimes it is hilariously sublime and it manages to capture enough 'soul' that it turns out a really creative image. Overall though, after a while it gets kinda same-y, and some things it just can't manage although I've had some luck with Sphinxes and Taurs.
Dominus tecum
Yes, AI is just fancy image manipulation. Even with ControlNet you still can't match a human artist especially since the only way you can get art that isn't fugly is by targeting an artist(s) in the prompt itself. It's a great inspiration tool though! I use StableDiffusion locally for fun.
Every time there is a new art technology people panic like when stock image databases like Stutterstock. Or DAZ 3D for 3d models.
If you'd like to play with some free AI here is a story writing one and an image one. Make sure to turn off adblock so it doesn't refresh the page and be warned that they are both uncensored so you might get adult work. There are more on the site but these are the main ones.
https://perchance.org/ai-story-generator
https://perchance.org/ai-text-to-image-generator
In a way, that sorta gives us old traditional media art folks a one-of-a-kind advantage for now.