Generative AI Art

[Edenray]

Okay, thank you for being clear. I think what I will do is remove the imagery for an upload, and email it if anyone wants to help complete the artwork with me. The other thing I could do for now is just to sketch out basic outlines of what I want to upload, not copying the AI images as they are literally just idea guidelines.
Thanks again!

[bssarkar]

I fail to understand what's ambiguous about this: https://openai.com/policies/row-terms-of-use/

Ownership of content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output.

The original post forgot to or left vague a very important thing: Money

All donations or a part of it goes to Wesnoth mainline art comission, like the portraits by Kitty/LordBob. Mainline can afford that. But can UMC? They have as much right to tell a story as has mainline, do they not?
On the reverse side: the artists do suffer since they are the one who would be losing their commissions first.
Also, Wesnoth is a FOSS game: we don't try to actively make money by using microtransaction or something like that. However, money still goes out, does it not? Current SP revisions have already commissioned artwork by now I think.
Now, this point is useless if artists did provide free images, but they probably can't afford it AFAIK.

the art novice branch would be lost.

Show me someone who can create LordBob level art in UMC and who is currently active? Most people are simply creating pixel art and not portrait.

We would gain some assets, but we would lose some of our community function and we would be embracing the framework of immoral people.

How do you know they are immoral?

More vague points:
1. What if somebody generates reference photos and use them to draw a picture by his own?
2. AI's are advanced or will advance to the point where they probably can copy LordBob's or Kitty's style so closely that they are indistinguishable. What if somebody sneaks that in? How'd you prove that's not done by him/her?

I'd admit though that AI generated art is still a legally grey area, so that could be a point against it.

We could probably add an indemnification clause to Wesnoth (if it isn't covered by GPL already) that any content's sole responsibility is it's owners'. This way, we don't face lawsuits and can remove that content if somebody warns us. [obvious disclaimer: I'm not a law expert. Best to consult a lawyer on the finer points, but there should be a way out.]

[Note: I don't hold a personal grudge towards anyone.]

[bssarkar]

The previous post tells my arguments. ChatGPT has it's own arguments too (XD), which interested people can read here: (public link)
https://chatgpt.com/share/6727755b-0c54 ... df772e7781

[Note: once again, I don't want to attack or offend anyone. Simply debating.]

[gnombat]

I fail to understand what's ambiguous about this: https://openai.com/policies/row-terms-of-use/

Ownership of content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output.

The problem is that OpenAI is likely using third-party artwork (as input to their AI) that they do not have any rights to use.

OpenAI can assign away "all our [meaning OpenAI's] right, title, and interest" to you, but they can't assign away the rights of third parties (who never agreed for their work to be used like this in the first place).

the art novice branch would be lost.

Show me someone who can create LordBob level art in UMC and who is currently active? Most people are simply creating pixel art and not portrait.

I think that's the point - most novice artists simply don't have LordBob's level of skill. But everyone has to start somewhere, right? I think it's best if there is a pathway open for anyone to contribute to the game, regardless of skill level. Suppose that you're a novice artist, and you want to contribute to the game, but your art isn't up to mainline quality standards. You can still contribute art to UMC - it's possible that some people might even be willing to pay you a commission for it (although you probably wouldn't make as much money as LordBob would).

If everyone just starts using AI-generated artwork for their UMC, then that pathway is gone.

AI's are advanced or will advance to the point where they probably can copy LordBob's or Kitty's style so closely that they are indistinguishable. What if somebody sneaks that in? How'd you prove that's not done by him/her?

Well, I don't think it is possible to enforce 100% compliance with these rules - I think the hope is that most UMC authors will simply voluntarily comply with them. But of course the Wesnoth project might take action if it sees any evidence that these rules have been violated. For example, if an add-on contains a portrait that strangely has too many fingers...

[bssarkar]

The problem is that OpenAI is likely using third-party artwork (as input to their AI) that they do not have any rights to use.

OpenAI can assign away "all our [meaning OpenAI's] right, title, and interest" to you, but they can't assign away the rights of third parties (who never agreed for their work to be used like this in the first place).

Yeah, this legal grey area is probably the biggest objection to the widespread adoption.

I think that's the point - most novice artists simply don't have LordBob's level of skill. But everyone has to start somewhere, right? I think it's best if there is a pathway open for anyone to contribute to the game, regardless of skill level. Suppose that you're a novice artist, and you want to contribute to the game, but your art isn't up to mainline quality standards. You can still contribute art to UMC - it's possible that some people might even be willing to pay you a commission for it (although you probably wouldn't make as much money as LordBob would).

If everyone just starts using AI-generated artwork for their UMC, then that pathway is gone.

Both of these I think depends on the Add-on's creator. The person more likely to use a AI-generated art would also be the same person more likely to reject low quality art. On the other hand, the exactly opposite person would give a novice artist a chance even if AI art was an option, although the temptation to use AI art would be higher. From a story writer POV, low quality art simply destroys the Suspension of Disbelief in the story, resulting in lesser chances of actually creating a high quality campaign, which in turn hurts Wesnoth's capability to compete with other games in the market. That in turn causes lesser chances of getting new players and more chances of Wesnoth getting extinct in the near future.
The extremely high quality of art in mainline also sets the bar too high, even if it's UMC, because one look there and it sorta triggers the inferiority complex in the artist, resulting in much less chance of actually getting an artwork contribution. How many pixel artists are there compared to the number of the portrait artists?

[gnombat]

Both of these I think depends on the Add-on's creator. The person more likely to use a AI-generated art would also be the same person more likely to reject low quality art. On the other hand, the exactly opposite person would give a novice artist a chance even if AI art was an option, although the temptation to use AI art would be higher. From a story writer POV, low quality art simply destroys the Suspension of Disbelief in the story, resulting in lesser chances of actually creating a high quality campaign, which in turn hurts Wesnoth's capability to compete with other games in the market. That in turn causes lesser chances of getting new players and more chances of Wesnoth getting extinct in the near future.

Well, another issue is that the AI images themselves have an odd, soulless quality about them (not to mention sometimes the wrong number of fingers) which can also interfere with suspension of disbelief. A lot of the images in the original AI-generated art topic are like this.

Of course, AI is improving over time and is likely to be a lot better 5 or 10 years from now. So possibly it may be necessary to revisit this issue in the future. But right now AI-generated art seems pretty dubious in terms of quality, morality, and legality.

[bssarkar]

Well, another issue is that the AI images themselves have an odd, soulless quality about them (not to mention sometimes the wrong number of fingers) which can also interfere with suspension of disbelief. A lot of the images in the original AI-generated art topic are like this.

Of course, AI is improving over time and is likely to be a lot better 5 or 10 years from now. So possibly it may be necessary to revisit this issue in the future. But right now AI-generated art seems pretty dubious in terms of quality, morality, and legality.

I never said we should use AI images as-is. We could just get them edited by a (novice) human artist, which is much easier than create something from scratch. Or we could upload a rough sketch by an new artist, using that as base for AI generation, which could also resolve the dilemma somewhat. Also depends on the prompt and the AI engine obviously.

But right now AI-generated art seems pretty dubious in terms of quality, morality, and legality.

Dubious only in terms of legality. Dubious in terms of morality in the sense that artists lose their earnings.

Attaching an image as a sample, generated currently using ChatGPT4 free tier, along with the corresponding prompt. ChatGPT4 also uses DALL-E as backend (most probably DALL-E3). (I'm assuming I'm allowed to pick among all images generated by the AI used.)

draw a sad scene of an elf lord silently walking away from a huge tree which was planted in honour of his beloved deaf wife. Inside a jungle.

https://chatgpt.com/share/6727b21a-cbac ... 94da4f9f85

[DuncanDill]

That picture lacks soul so much...

[gnombat]

That picture lacks soul so much...

For AI, it's not bad (the elf even has the correct number of fingers). Personally, though, I would rather see a work which was created by a human being (even if it's not someone of LordBob's caliber).

[Pentarctagon]

I fail to understand what's ambiguous about this: https://openai.com/policies/row-terms-of-use/

Ownership of content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output.

The legal grey area likewise means it's legally dubious that OpenAI can actually make such a statement in the first place.

[ZombieKnight]

Well, another issue is that the AI images themselves have an odd, soulless quality about them (not to mention sometimes the wrong number of fingers) which can also interfere with suspension of disbelief. A lot of the images in the original AI-generated art topic are like this.

Of course, AI is improving over time and is likely to be a lot better 5 or 10 years from now. So possibly it may be necessary to revisit this issue in the future. But right now AI-generated art seems pretty dubious in terms of quality, morality, and legality.

To say it from my lens:
Because existing story images don't cover everything and if one's not an artist, has no reputation, and doesn't want to pay an artist to get his own story images (I go into this category too), the best way to get some story art is through AI.
(I'm absolutely not saying that they are perfect or better than the artwork of real artists, but it's better than nothing)

[bssarkar]

Still more questions:
1. What if someone uses an AI generated image as a reference image and draws the image themselves?
2. What about using AI generated art as a base image and then editing it?
3. What about taking AI advice about improving art skills/learning about techniques/tools?
4. If we're so friendly to novice artists, why does mainline keeps commissioning artwork? Why not take low quality artwork from new artists? Why is the bar so high? I have a suspicion that if one of the artist from Wesnoth's past changes their username gives the same level of art as was in say 1.4, then it probably won't get accepted. We say we're willing to live with low quality artwork just because it's human, but isn't this a double standard?
5. Once again, I fail to see this Soulless quality in AI artwork, especially in newer engines. And even if it is, I have a feeling that some people would still be willing to take it. While others won't. That's a personal choice. We could have a review panel if necessary to check if the generated images if necessary.

[Simply making a point here, attacking someone is not intended. I mean no offense to anyone, and apologize in advance.]

Another image (non-Wesnoth, but from my public archives of AI generated art from Leonardo AI, model Anime XL):
https://cdn.leonardo.ai/users/88d352f2- ... n_wo_2.jpg

This one has more contrast and I don't see anything soulless either.

[bssarkar]

What happens when I feed back the previous DALL-E generated elf lord scene into Leonardo:
https://cdn.leonardo.ai/users/88d352f2- ... _elf_1.jpg

The eyes are fixed for example and it does have correct hand anatomy, and more contrast.

[DuncanDill]

Im not entirely sure what it is, but it just looks bad. Something about it is wrong, there is no love or soul. If ai is accepted I might just go back to singleplayer. I really hate the look of it...

[ZombieKnight]

Im not entirely sure what it is, but it just looks bad. Something about it is wrong, there is no love or soul. If ai is accepted I might just go back to singleplayer. I really hate the look of it...

Man your arguments are far from valid, you look on image and say it has no love... I can assure you that my story image would have love, but would look much worse.
-_-