FA+
Shop
Some thoughts on Ai art generators
a year ago
I'm a writer at heart, and to prove it I've written a growing series of books, featured within the Ephalon universe of my creation.
Ephalon website
Ephalon - book purchase page
Ephalon Twitter page I've been practising with a few ai art generators in my off time to get an idea of what this technology can do.
Here are some bullet points which I think artists and commissioners should keep in mind.
For context: I'm coming at this topic from a realistic practical angle. I'm not interested in the morality or the ethics of these things.
1) It's very obvious that the tech mixes and matches similar/desired art pieces together to create the prompted image. To that end, it functionally can't create something original; here's what I mean.
To truly compare, you need to analyse the method in which a human mind creates art and this computer software.
- A human mind abstracts ideas from real world into concepts and then melds them together to draw something new. You need some visual input of course, but sometimes people can create something from nothing as well.
- The ai art generator needs raw input data with knowledge of the pixel patterns its seeing and how with certain elements it should interact in the realm of 2d artwork.
If you read both of those things you should realise they are not the same. And I realised that exceptionally complex poses, or ones that are not possible in reality, but are possible to draw, are not possible through ai artwork, it will ALWAYS need data input to base it on.
So, artists who are good at rare/complex posing, you're good. Ai generators can't do this (well/consistently), and likely will never be able to, simply due to a lack of data.
2) AI generators, aren't really AI. That's a buzz term. What it actually is is a complex algorithm that mixes and matches things based on input patterns and associated key words.
3) I believe that down the line, AI art generators are likely(highly probable) to consume so much data(artwork) produced by themselves, which will inevitably lead to quality dilution and/or degradation over time.
4) So far I've not seen any ai art generators that adhere to real world concepts such a physics and biology in real-time. That's not to say this might not one day become a thing, but there's an issue there. A normal artist of high quality knows intrinsically how something 'should' look in an image purely based on biology and physics. This machine pattern recognition algorithm, so far I've seen throughout my experiments, only knows preferred patterns.
Practical consequences as I see them for artists:
- I expect that low end buyers will be less likely in procuring artwork for role play purposes.
- Good (skilled) artists, will not be affected.
- Artists who do rare or complex poses, will not be affected.
- Unique, though not highly popular styles should be fairly unaffected as well.
- Commissioners who approach an artist and then then use the excuse it's too expensive I can get it AI generated anyway or for less through that method, weren't interested in buying a picture from you to begin with. It's rude and just ignore these people.
- The adopt market I think will suffer greatly.
- Ownership will be an issue. If you commission something that image is effectively yours. If you Ai generate something, it's a mixture of the work of those who came before you. It's murky territory and I believe there are ongoing legal battles about this, which may have industry wide ramifications. I look forward to the conclusions.
- There are new opportunities for artists to draw basic stuff to be used for ai generators. I've not seen this being sold as a product yet; what do I mean with this? Rare poses to use as base images. You could create a pose pack, with colour sliders, clothing etc. This could be huge if done right. If you as an artists take the time to learn this tech, you can see how there's a new market for different kinds of artwork to be made. A mixture of commission work for the purpose of ai artwork.
These are some thoughts I've collected since using the generator.
I've seen all kinds of discussion about the ethics.
I've seen all kinds of artists become nihilistic about art.
Yet, I've rarely seen people jump on this to find new art markets, or opportunities.
You may not agree with it, or maybe you love it. But Pandora's box has been opened, and the world has changed forever. It's better to accept and adapt.
But rest assured normal artists will still be needed, since as I mentioned above, true originality cannot be created by an algorithm.
On side note, if anyone is experiencing a downturn in commission interest, please be aware the world is effectively in a recession with endlessly rising costs. Sadly, artwork which isn't food, heating etc. Isn't a priority at that point. I believe that's the most likely cause.
Here are some bullet points which I think artists and commissioners should keep in mind.
For context: I'm coming at this topic from a realistic practical angle. I'm not interested in the morality or the ethics of these things.
1) It's very obvious that the tech mixes and matches similar/desired art pieces together to create the prompted image. To that end, it functionally can't create something original; here's what I mean.
To truly compare, you need to analyse the method in which a human mind creates art and this computer software.
- A human mind abstracts ideas from real world into concepts and then melds them together to draw something new. You need some visual input of course, but sometimes people can create something from nothing as well.
- The ai art generator needs raw input data with knowledge of the pixel patterns its seeing and how with certain elements it should interact in the realm of 2d artwork.
If you read both of those things you should realise they are not the same. And I realised that exceptionally complex poses, or ones that are not possible in reality, but are possible to draw, are not possible through ai artwork, it will ALWAYS need data input to base it on.
So, artists who are good at rare/complex posing, you're good. Ai generators can't do this (well/consistently), and likely will never be able to, simply due to a lack of data.
2) AI generators, aren't really AI. That's a buzz term. What it actually is is a complex algorithm that mixes and matches things based on input patterns and associated key words.
3) I believe that down the line, AI art generators are likely(highly probable) to consume so much data(artwork) produced by themselves, which will inevitably lead to quality dilution and/or degradation over time.
4) So far I've not seen any ai art generators that adhere to real world concepts such a physics and biology in real-time. That's not to say this might not one day become a thing, but there's an issue there. A normal artist of high quality knows intrinsically how something 'should' look in an image purely based on biology and physics. This machine pattern recognition algorithm, so far I've seen throughout my experiments, only knows preferred patterns.
Practical consequences as I see them for artists:
- I expect that low end buyers will be less likely in procuring artwork for role play purposes.
- Good (skilled) artists, will not be affected.
- Artists who do rare or complex poses, will not be affected.
- Unique, though not highly popular styles should be fairly unaffected as well.
- Commissioners who approach an artist and then then use the excuse it's too expensive I can get it AI generated anyway or for less through that method, weren't interested in buying a picture from you to begin with. It's rude and just ignore these people.
- The adopt market I think will suffer greatly.
- Ownership will be an issue. If you commission something that image is effectively yours. If you Ai generate something, it's a mixture of the work of those who came before you. It's murky territory and I believe there are ongoing legal battles about this, which may have industry wide ramifications. I look forward to the conclusions.
- There are new opportunities for artists to draw basic stuff to be used for ai generators. I've not seen this being sold as a product yet; what do I mean with this? Rare poses to use as base images. You could create a pose pack, with colour sliders, clothing etc. This could be huge if done right. If you as an artists take the time to learn this tech, you can see how there's a new market for different kinds of artwork to be made. A mixture of commission work for the purpose of ai artwork.
These are some thoughts I've collected since using the generator.
I've seen all kinds of discussion about the ethics.
I've seen all kinds of artists become nihilistic about art.
Yet, I've rarely seen people jump on this to find new art markets, or opportunities.
You may not agree with it, or maybe you love it. But Pandora's box has been opened, and the world has changed forever. It's better to accept and adapt.
But rest assured normal artists will still be needed, since as I mentioned above, true originality cannot be created by an algorithm.
On side note, if anyone is experiencing a downturn in commission interest, please be aware the world is effectively in a recession with endlessly rising costs. Sadly, artwork which isn't food, heating etc. Isn't a priority at that point. I believe that's the most likely cause.
2) This is correct. Lots of people think AI image generators are going out and searching for images and are currently learning, but they're not. It just references its massive model size that basically tells it "when this word, then this type of pixel arrangement".
3) I've seen this argument going around, but there's nothing stopping people from just using older models indefinitely. People just won't use new models if they're clearly poisoned enough to not give better results. Right now, if you're SUPER good at prompting (yes, it's a skill, and no I'm not the best at it), you can get genuinely amazing results, to the point where a lot of people would argue that they wouldn't even need to download new models ever again.
As for me, I see it as a tool that can help me get some quick concept art that I can then take to a real artist to draw out properly. It's also a tool that artists can use to improve their own pipeline. Ultimately it requires creativity to actually be good, and that requires an artistic eye to some degree, otherwise it's just samey slop (which I'll admit is a great chunk of AI generated content).
Artists who draw unique niche sorts of things and have a specialty and do it well are also going to be unaffected.
- Good (skilled) artists, will not be affected.
- Artists who do rare or complex poses, will not be affected.
Are completely false. Unless you've narrowed your sense of who you care about to 'furry commission-exclusive artists', rather than the 99% of artists who don't do that.