Post something that you worked on recently!

[sootesheepy666]

A quick half hour doodle because I finally found some time to draw something.

KAngle!! PEAK

 

[sootesheepy666]

Ai made. My own work. Yadda yadda yadda. I did not use a reference image.

How is it your own work if AI made it? You only had to write a prompt

 

[IlaiaCorde]

How is it your own work if AI made it? You only had to write a prompt

Well first off,

Thats complete nonsense. Its the typical kind of thing that people who dont understand the process usually say, because they dont even understand how the technology works.

I actually drew the sketch that created that image (with no initial reference for the sketch itself). The AI turned that sketch into the image above. The idea it all comes from a prompt in every instance is incorrect. You CAN make an image entirely from a prompt (or even without one), but thats not always the case. It depends on the user.

Initial sketch:



But I still want to address your criticism. Because I dont think it would matter even if all I did was type a prompt. Lets talk about the steps that go into it:

0.) First and most important is artistic vision. If you can't imagine it, you can't create it. Although you can get some random results if you type nothing at all (in some generators). Or slightly less random results if you type something simple, like a sentence or two.

1.) Writing the prompt is pretty crucial. Although in fact its still only about 50% of what the initial results will be. And just to be clear, most of the prompts required to get something this specific are usually multiple paragraphs in length and require some technical understanding to pull it off. But either way...if it wouldn't have existed without the prompt, then that sounds to me like its "mine" since I wrote the instructions that created that specific result.

Would it have existed if I had done nothing? Or done something different? No. It wouldn't. So if it exists solely by my actions in a process...how is that not mine?

If a story written by a writer/author is theirs because they typed it, how would this be any different? Answer...its not. I had an idea and used a tool to bring it to life; that certainly sounds like its mine.

2.) Its actually not just a prompt. It also requires upscaler settings, loras, HD settings, shader settings, lighting settings, rendering settings ect. Some of it is configurable via drop down menus, others require typing it into the prompt itself. Technical know how is important for AI since AI is essentially a kind of software.

3.) Post-effects are pretty important. The idea that genuinely good 'AI' artists just post whatever comes out of the generator is misguided. I mean sure...some do just that. But if you want something really specific, its going to require more work. Maybe a lot more.

4.) Hand editing. Sometimes the artist has to edit the photo by hand. Although AI error isnt as bad as it was in the early days. This requires some actual artistic ability, even if its just removing or add a minor detail or feature.

5.) AI art isnt always entirely AI generated. Many programs like Adobe'a art programs have generative fill features. This varies in how much it can so. Some will jist fill in a selected area. Like you dont want to draw the hair...so you select the area you want the AI to fill, and type in a prompt for what you want in that space (in this case hair, and again, you have to be specific) and it will fill that spot in for you.

But it can go way beyond that. Since now we can generate off a sketch or an initial hand created reference image. Meaning you could draw a sketch on paper, upload it to phone or pc, and then into the generator and the AI will take that sketch and turn it into a full artwork based entirely on the sketch. Thats what I did here.

‐---------------

So this idea that all AI art is 100% AI generated with no other input is really misinformation. Its also foolish to think that just because an AI Artist is using AI, they cant draw or do any 'by hand' art. Many can, and use AI to streamline the process. Some of us do both (like me).

Most importantly...do you think...that art can only be a product that comes from humans who learned to create something using motor / mechanical skills? I find that weird. Why would it matter? Why is a pencil or a stylus the only tool allowed? If a keyboard or a mouse can accomplish the same task but with less hassle? A pencil is just a tool. So is a keyboard. So is an AI. An AI is just a tool.

Its a tool that is a million times more versatile and more efficient than a pencil.

 

[sootesheepy666]

Well first off,

Thats complete nonsense. Its the typical kind of thing that people who dont understand the process usually say, because they dont even understand how the technology works. 🤣

I actually drew the sketch that created that image (with no initial reference for the sketch itself). The AI turned that sketch into the image above. The idea it all comes from a prompt in every instance is incorrect. You CAN make an image entirely from a prompt (or even without one), but thats not always the case. It depends on the user.

Initial sketch:



But I still want to address your criticism. Because I dont think it would matter even if all I did was type a prompt. Lets talk about the steps that go into it:

0.) First and most important is artistic vision. If you can't imagine it, you can't create it. Although you can get some random results if you type nothing at all (in some generators). Or slightly less random results if you type something simple, like a sentence or two.

1.) Writing the prompt is pretty crucial. Although in fact its still only about 50% of what the initial results will be. And just to be clear, most of the prompts required to get something this specific are usually multiple paragraphs in length and require some technical understanding to pull it off. But either way...if it wouldn't have existed without the prompt, then that sounds to me like its "mine" since I wrote the instructions that created that specific result.

Would it have existed if I had done nothing? Or done something different? No. It wouldn't. So if it exists solely by my actions in a process...how is that not mine?

If a story written by a writer/author is theirs because they typed it, how would this be any different? Answer...its not. I had an idea and used a tool to bring it to life; that certainly sounds like its mine.

2.) Its actually not just a prompt. It also requires upscaler settings, loras, HD settings, shader settings, lighting settings, rendering settings ect. Some of it is configurable via drop down menus, others require typing it into the prompt itself. Technical know how is important for AI since AI is essentially a kind of software.

3.) Post-effects are pretty important. The idea that genuinely good 'AI' artists just post whatever comes out of the generator is misguided. I mean sure...some do just that. But if you want something really specific, its going to require more work. Maybe a lot more.

4.) Hand editing. Sometimes the artist has to edit the photo by hand. Although AI error isnt as bad as it was in the early days. This requires some actual artistic ability, even if its just removing or add a minor detail or feature.

5.) AI art isnt always entirely AI generated. Many programs like Adobe'a art programs have generative fill features. This varies in how much it can so. Some will jist fill in a selected area. Like you dont want to draw the hair...so you select the area you want the AI to fill, and type in a prompt for what you want in that space (in this case hair, and again, you have to be specific) and it will fill that spot in for you.

But it can go way beyond that. Since now we can generate off a sketch or an initial hand created reference image. Meaning you could draw a sketch on paper, upload it to phone or pc, and then into the generator and the AI will take that sketch and turn it into a full artwork based entirely on the sketch. Thats what I did here.

‐---------------

ISo this idea that all AI art is 100% AI generated with no other input is really misinformation. Its also foolish to think that just because an AI Artist is using AI, they cant draw or do any 'by hand' art. Many can, and use AI to streamline the process. Some of us do both (like me).

Most importantly...do you think...that art can only be a product that comes from humans who learned to create something using motor / mechanical skills? I find that weird. Why would it matter? Why is a pencil or a stylus the only tool allowed? If a keyboard or a mouse can accomplish the same task but with less hassle? A pencil is just a tool. So is a keyboard. So is an AI. An AI is just a tool.

Its a tool that is a million times more versatile and more efficient than a pencil.

In favor of polite debate, here is my opinion. I do not mind this form of AI nearly as much as normal since you did the sketch but normally, without premade sketches, the art is entirely composed of stolen art. The meaning and purpose of art is to connect others and show your emotion. Every line and brush stroke is deliberate. So, while AI art may be pretty sometimes, it is never emotional and soulful the way that real art is. I think AI should be used to take over menial jobs that no one likes, not something like music or art that is so wholly human.

 

[IlaiaCorde]

In favor of polite debate, here is my opinion. I do not mind this form of AI nearly as much as normal since you did the sketch but normally, without premade sketches, the art is entirely composed of stolen art. The meaning and purpose of art is to connect others and show your emotion. Every line and brush stroke is deliberate. So, while AI art may be pretty sometimes, it is never emotional and soulful the way that real art is. I think AI should be used to take over menial jobs that no one likes, not something like music or art that is so wholly human.

I disagree with that on its very premise.

Is it actually stolen art? No, it isn't. Thats a falsehood. Its been discussed numerous times in many different contexts.

How does a diffusion model for instance learn to make images?:

Specifically-

Diffusion Models (e.g., Stable Diffusion, DALL-E 3): These are currently the most popular for high-quality results. The model is trained to "denoise" images. During training:
Start with a real image from the dataset.
Gradually add random noise (like static on a TV) until it's pure randomness.

The AI learns to reverse this: predicting and removing noise step by step to reconstruct the original image.
Text prompts are incorporated using a separate component (like CLIP) that encodes descriptions into numerical representations (embeddings), guiding the denoising toward matching concepts.

Over billions of iterations on powerful GPUs, the model improves by minimizing errors, essentially learning statistical patterns in pixel distributions rather than storing copies.

When it understands, it can come up with things entirely on its own.

Specifically once trained, the AI doesn't need the full dataset anymore—it uses the learned patterns. Here's the inference process:

You input a text prompt (e.g., "a cyberpunk city at night").
The text is encoded into embeddings that capture semantics.

Starting from random noise (for diffusion) or a seed, the model iteratively refines it: In diffusion, it denoises over 20–100 steps, guided by the prompt, to build coherent pixels.
Output: A new image that's a novel combination of learned elements, not a direct copy from training data.

So its not tracing or borrowing pixels from existing works. And, it could be using actual photographs to understand concepts. For example, if you typed 'yellow jacket' into the prompt, the AI could be looking at hundreds of photos of yellow jackets (the clothing, not the insect) on/from Google Images and determining what that looks like. It doesnt need a yellow jacket from an anime image to then draw a yellow jacket in an anime style. It can do it from scratch.

It could be using images from retail websites for instance, perhaps in google images, to understand what yellow jackets look like.

In other words, it meets none of the qualifications for art plagiarism. Which is the ONLY real metric for determining whether art is stolen, outside of legal access (which is still a legit issue).

-------

So now we understand that, it comes down to whether they had permission to 'use' existing artwork in their training data.

Theres several issues at play here. Some of which are being handled in courts of law still.

But ultimately it comes down to whether a machine has the right to view an image. This is called Fair Use in the legal context its being used.

The AI doesnt need to store it. It only needs to view it repeatedly. The question then becomes...did they access the image legally? Pirated literature has been ruled not fair use in the ongoing LLM litigation. That will probably be true for Image AIs.

-----

Do humans go through a similar process when learning to create art?

Probably. We don't understand the human consciousness yet really, or even how the brain works to produce consciousness, so its impossible to say for sure.

But what we can say, is that no human is born from the womb making art. It has to be learned. Nd even then, its going to be based on what the artist has learned. I.e, reproducing things they have seen. Or things they can imagine, which are heavily influenced from what they've seen and experienced.

Humans dont require permission to do this. They see. They learn. They remember. They understand.

So what this all boils down to, is does a machine have a right to view and understand something thats in public view? I.e, images on the internet? The bottom line is...that if its not illegal for a human to look at an image and learn from it, its not illegal for a machine either. Let alone a machine being operated by humans.

What we should be concerned with, is whether the images were accessed legally. Thats the big issue.

------

Lastly.

But most important. This is really going to be the death blow to the Anti Ai arguments in court (outside of whether images were accessed legally. Thats still in play): Can AIs create or draw artworks based on actual photographs?

Yes. So if you had a photo of a subway station, you could show it to an image AI and ask for it to make an anime art reproduction of the image. And it could do it.

So they can draw or create things they see. This is a big deal, because the AI doesnt need our artwork nearly as much as Anti AI people claim. The claims are mostly falsehoods.

Image AIs mostly require actual artwork simply to understand genre, and stylistic choices. I.e, a subway train looks this qay in reality (in a photograph). But in anime art, it usually looks this other way.

This issue is why...AIs can absolutely produce entirely new artwork. Its fully possible. The idea AIs can only reproduce what they've seen is incorrect. They can combine ideas, styles and concepts to create entirely new things that wasn't seen previously in their training data.

------

Does AI have soul?

You made this argument above and said it does not.

I disagree. AI embodies the soul and ideas of the human who uses it. So it does have soul. And it certainly is meaningful.

As always, art is subjective. People will probably never agree on an exact definition of what is art, and what is not art.

Thats why I simply use the actual definition:

art /ärt/

noun

The conscious use of the imagination in the production of objects intended to be contemplated or appreciated as beautiful, as in the arrangement of forms, sounds, or words.

Such activity in the visual or plastic arts.
"takes classes in art at the college."

Products of this activity; imaginative works considered as a group.
"art on display in the lobby."

-------

Does AI qualify for that? If a human is operating it, yes.

And thats the way I see it.

 

[IlaiaCorde]

One other thing...

I dont think AI art is stolen art at all (And I think thats a proven fact).

But the thing that I find truly bizarre...is that people on this site would even be saying such a thing here on Mangadex in the first place. Mangadex is a site that is hosting copyrighted works without the author/artists permission. The "largest manga piracy site in the world" according to publishers and media. (I dont think they are right about that btw)

Meanwhile, people are claiming a moral high ground on AI art (which is actually completely legal)? The weird flex going on here on the forum is super weird.

Im not trying to offend anyone here, but theres definitely a lack of situational-awareness with this that is pretty mind boggling.

 

[Madologist]

One other thing...

I dont think AI art is stolen art at all (And I think thats a proven fact).

But the thing that I find truly bizarre...is that people on this site would even be saying such a thing here on Mangadex in the first place. Mangadex is a site that is hosting copyrighted works without the author/artists permission. The "largest manga piracy site in the world" according to publishers and media. (I dont think they are right about that btw)

Meanwhile, people are claiming a moral high ground on AI art (which is actually completely legal)? The weird flex going on here on the forum is super weird.

Im not trying to offend anyone here, but theres definitely a lack of situational-awareness with this that is pretty mind boggling.

Take it here mayhap?

 

[Manko-sensei]

And this is what happens when people don't read Kant's critiques.

 

[Madologist]

Well back on topic, guess I'll post my most recent overmap made with Wonderdraft.

 

[IlaiaCorde]

Take it here mayhap?

Yes.

Probably a better thread for it.

Im not the one that started the debate here. I was really just responding to someone else and I didn't feel comfortable telling then to switch threads. 😅

Well back on topic, guess I'll post my most recent overmap made with Wonderdraft.

I actually really like that. I make maps sometimes too (by hand).

Has an old world feel to it.

 

[IlaiaCorde]

And this is what happens when people don't read Kant's critiques.

I actually have.

Is there something specific you think is relevant here?

For instance, the idea that AI’s role in creativity may deepen Kant’s idea that beauty and meaning are co-produced by perceivers, not properties of objects?

Still...I dont find that relevant. I dont think AI decides what is beautiful. I think the person using it does. Ai operates under the users purview, limitations in that will naturally limit the usefulness of AI as a tool.

If you wish to respond, please use the thread Madologist suggested:

https://forums.mangadex.org/threads/artist-vs-ai.1603323/

Just tag me in with an @ username and I will respond when I get some time.

 

[Madologist]

I actually really like that. I make maps sometimes too (by hand).

Has an old world feel to it.

Wonderdraft. Use premade assets to play with legos, but its maps. You can slap in anything as assets, which is very useful, since sometimes you do have to make them yourself, when more talented people havn't delivered (for free). Same folk have made Dungeondraft for battle maps too.

 

[flxsemi]